Publications

In this study, the researchers aimed to evaluate the accuracy and correlation between an artificial intelligence (AI) software, Coreline AVIEW from Seoul, South Korea, and expert human interpretation in assessing coronary artery calcium (CAC) scores using non-contrast calcium score images. A total of 100 studies were randomly selected and processed by both the AI software and human level-3 computed tomography (CT) reading. The results demonstrated a highly significant correlation between the AI software and human reading in terms of absolute CAC scores. However, despite minimal score differences, 14% of the patients experienced reclassification in the coronary artery disease data and reporting system (CAC-DRS) category. The main cause of reclassification was observed in the CAC-DRS 0-1 category, where misclassification was influenced by factors such as AI underestimation of right coronary calcium, AI overestimation of right ventricle densities, and human underestimation of right coronary artery calcium. Overall, the study highlights the strong correlation between AI and human values in CAC scoring, with the need for further algorithm optimization to improve AI sensitivity and specificity, particularly for minimal disease detection. The AI software showed excellent correlation with human expert reading across a broad range of calcium scores and occasionally identified calcium that was missed by human interpretation.

The present analysis investigates the predictive capacity of a fully automated coronary artery calcium (CAC) scoring system using artificial intelligence (AI) software (AVIEW, Coreline Soft, Seoul, Korea) in a low-dose computed tomography (LDCT)-based lung cancer screening (LCS) trial. The study includes 2239 participants from the Multicentric Italian Lung Detection (MILD) trial who underwent a baseline LDCT between September 2005 and January 2011, with a median follow-up of 190 months. The CAC scores were categorized into five strata, and their association with 12-year all-cause mortality and non-cancer mortality was analyzed. The results show that higher CAC scores, particularly CAC > 400, are associated with increased all-cause mortality and non-cancer mortality rates. The fully automated CAC scoring system effectively predicts all-cause mortality at 12 years in the LCS setting. These findings highlight the potential of automated CAC scoring in enhancing risk assessment and mortality prediction in LDCT-based LCS.

The prevalence of interstitial lung abnormalities (ILAs) in the Korean lung cancer screening population was found to be 4%, as identified in this retrospective study. The study aimed to quantitatively assess ILA with CT examinations from the Korean National Lung Cancer Screening Program using a deep learning–based texture analysis. The study included participants who underwent chest CT between April 2017 and December 2020. The findings were classified by three radiologists into three groups: no ILA, equivocal ILA, and ILA (fibrotic and nonfibrotic). ILA progression was evaluated between baseline and last follow-up CT scans. The extent of ILA was assessed visually and quantitatively with the help of deep learning–based texture analysis using the AVIEW Lung Texture ILA, v 1.1.39.14 software by Coreline Soft. This deep learning-based lung texture analysis software demonstrated high sensitivity and specificity for detecting ILA with a 1.8% lung area cutoff value. This study indicates that ILAs are prevalent in the Korean lung cancer screening population and can be effectively detected and quantified using deep learning–based texture analysis.

In this retrospective study, a simplified ordinal scoring method referred to as the modified length-based grading was validated for assessing coronary artery calcium (CAC) severity on non-electrocardiogram (ECG)-gated chest computed tomography (CT). The study included 120 patients who underwent both non-ECG-gated chest CT and ECG-gated cardiac CT. Six radiologists independently used two scoring methods, visual assessment and modified length-based grading, to assess CAC severity on chest CT. The Agatston score categories on cardiac CT served as the reference standard. The results indicated that the modified length-based grading method demonstrated better interobserver agreement and agreement with cardiac CT than visual assessment. In this study, CT images and patient information were reviewed and anonymized using a dedicated server, AVIEW Research, by Coreline Soft Inc., which was accessed by the six observers for image review and grading. This study concludes that the modified length-based grading method effectively evaluates CAC on non-ECG-gated chest CT.

In this study, researchers evaluated the pulmonary vasculature of 30 chronic thromboembolic pulmonary hypertension (CTEPH) patients using non-contrast computed tomography (CT) before and after treatment. They correlated CT-based parameters with right heart catheterization (RHC) hemodynamic and clinical parameters. Leveraging the Aview system (Coreline Soft, Inc.) for lung segmentation and pulmonary vasculature analysis, the researchers found significant increases in the number, area, and density of subpleural small vessels post-treatment. Additionally, blood volume shifted from larger to smaller vessels, and the BV5/TBV ratio correlated with hemodynamic and clinical parameters. The study highlights the effectiveness of non-contrast CT measures, aided by the Aview software, in evaluating changes in pulmonary vasculature in response to treatment.

This study investigated the prognostic value of deep learning (DL)-driven CT fibrosis quantification (using Coreline Soft AVIEW Lung Texture) in idiopathic pulmonary fibrosis (IPF). A total of 161 patients were evaluated, and CT-Norm% and CT-Fib% demonstrated significant correlations with forced vital capacity (FVC) and diffusion capacity of carbon monoxide (DLCO). Both CT-Norm% and CT-Fib% were found to be independent prognostic factors for overall survival in IPF when adjusted for various factors. The study concluded that CT-Norm% and CT-Fib% calculated using chest CT-based deep learning software could serve as reliable prognostic factors for IPF patients.

This study aimed to assess the use of automated coronary artery calcium (CAC) and quantitative emphysema (percentage of low attenuation areas [%LAA]) in predicting mortality and lung cancer incidence in lung cancer screening. Additionally, the study explored the correlations between %LAA, CAC, and forced expiratory value in 1 second (FEV1). The low-dose computed tomography (LDCT) images were analyzed using fully automated AI software, AVIEW, developed by Coreline Soft. A total of 4098 volunteers were enrolled, and the results showed that %LAA and CAC independently predicted 6-year all-cause, noncancer, and cardiovascular mortality. However, no significant association with lung cancer incidence was found after adjustments. Both biomarkers negatively correlated with FEV1, and %LAA demonstrated moderate discriminative ability in identifying airflow obstruction.

In this study, the position of the inferior alveolar nerve (IAN) was automatically imaged and tracked using artificial intelligence (AI) for safer surgery. The IAN segmentation was performed using AVIEW Modeler software (version 1.0.3, Coreline Software, Seoul, Korea), and the deep active learning framework proved to be a fast, accurate, and robust clinical tool for demarcating IAN location. Similarly, in another study, CT scans were analyzed using AVIEW Research (Coreline) software to measure abdominal fat depots in patients with obesity before and after bariatric surgery. The software was able to measure the volumes and fat attenuation index (FAI) of different fat depots accurately, and the study found that specific abdominal fat depot changes measured by CT were positively associated with metabolic health status conversion.

This study aimed to investigate the quantitative high-resolution computed tomography (HRCT) characteristics of Preserved Ratio Impaired Spirometry (PRISm) and evaluate the correlation between HRCT parameters and pulmonary function parameters. The study included 94 cases with normal pulmonary function, 51 cases of PRISm, and 48 cases of mild to moderate chronic obstructive lung disease (COPD). The lung parenchyma, parametric response mapping (PRM), airway, and vessel parameters were measured by automatic segmentation software (AVIEW, Coreline Soft). Quantitative CT parameters showed mild to moderate correlation with pulmonary function parameters. The results revealed significant differences in small airway and vessel lesions between PRISm and normal pulmonary function subjects, with PRISm more inclined to mild to moderate COPD. The nomogram based on quantitative HRCT parameters demonstrated good predictive value for the early screening of PRISm, providing more objective evidence.

AI, particularly deep learning, has become a significant area of research in thoracic oncology, benefiting thoracic imaging. This article reviews the current applications and perspectives of AI in this field. CADe tools have been available since the early 2000s for pulmonary nodule detection, with newer deep learning-based tools reducing false-positive results. Machine learning and deep learning methods have been employed for pulmonary nodule segmentation, volumetry, and characterization. AVIEW LCS was mentioned as an AI tool dedicated to lung cancer screening that can not only automatically detect and measure lung nodules but also suggest Lung-RADS categorization. Data from the NLST has led to the development of CADx tools for lung cancer diagnosis on chest CT scans. AI has also been used for virtual biopsies and predicting treatment response or survival. Numerous detection, characterization, and stratification tools have been proposed, with some being commercially available.

This study aimed to identify and quantify high coronary artery calcium (CAC) using deep learning (DL)-powered CAC scoring (CACS) in oncological patients with known very high CAC (≥ 1000) undergoing 18F-FDG-PET/CT for re-/staging. 100 patients were enrolled and divided into two groups. The fully automated DL-based CACS tool, AVIEW CAC (Coreline Soft, v1.1.42), was used to perform CACS on non-contrast, free-breathing, non-gated CT scans from 18F-FDG-PET/CT examinations. The results showed that the DL tool underestimated CAC load but correctly assigned an Agatston score ≥ 1000 in over 70% of cases, provided sufficient CT image quality. In the control group, the DL tool did not generate false-positives.

This study evaluated the performance of a fully automated quantitative software, AVIEW ILA (Coreline), in detecting interstitial lung abnormalities (ILA) on routine chest CT compared with radiologists' visual analysis. The study included 336 participants who underwent chest CT for health screening. Inter-reader agreements were substantial for the presence of ILA (weighted κ, 0.74) and fair for its subtypes (weighted κ, 0.38). The quantification system showed 67.6% sensitivity, 93.3% specificity, and 90.5% accuracy in identifying ILA using a threshold of 5% in at least one zone. The best cut-off value for detecting ILA was 3.6%. The study concluded that the inter-reader agreement was substantial for ILA but only fair for its subtypes, and using an automated quantification system in routine clinical practice may aid in the objective identification of ILA.

This study aimed to validate an AI-based automatic coronary artery calcium (CAC) scoring system on non-ECG-gated low-dose chest computed tomography (LDCT) using multi-institutional datasets. The AVIEW CAC software from Coreline Soft was used to perform automatic CAC scoring on LDCT scans, showing good to excellent reliability in CAC score and CAC severity categorization. However, the reliability varied among institutions. Overall, the AI-based CAC scoring system demonstrated promising results in detecting CAC on LDCT scans across multiple institutions.

Advances in artificial intelligence (AI), including deep learning, have significantly impacted medical imaging, leading to the development and commercialization of various AI algorithms for chest imaging. In this field, AI can be applied for initial evaluation, detection and diagnosis, quantitative assessment, monitoring, and decision support. The authors of this paper depicted several AI functions of AVIEW from Coreline Soft as good examples. However, some obstacles must be addressed for successful implementation. Familiarizing oneself with the current status, potential applications, and challenges of AI in chest imaging is essential for radiologists and clinicians in this evolving era.

The objective of this study was to determine the optimal method of defining growth for indeterminate pulmonary nodules detected in lung cancer screening. The study evaluated 115 nodules (93 solid and 22 part-solid) by measuring average diameters and volumes on baseline and follow-up LDCT scans using semi-automated segmentation. The results showed that percentage volume growth of ≥25% had higher sensitivity but lower specificity than diametrical measurement compared to absolute diameter growth of >1.5 mm. The radiologist's subjective decision had equivalent sensitivity but higher specificity compared to volume growth, but did not differ from diameter growth. Sensitivity and specificity of diagnostic referral based on volume doubling time (VDT) thresholds of ≤200 and ≤300 days were significantly lower and higher, respectively, than the VDT threshold of 600 days. A-view LungScreen, a commercial software from Coreline Soft, was used for measuring nodule size. The study suggests that growth evaluation of screening-detected indeterminate nodules with volumetric measurement exhibited higher sensitivity but lower specificity compared to diametric measurements.

The retrospective cohort study aimed to describe the incidence of atrial fibrillation (AF) in patients with lung cancer and to identify predictive cardiac dosimetric parameters. The study included 239 patients with limited-stage small cell lung cancer (SCLC) and 321 patients with locally advanced non-small cell lung cancer (NSCLC) who received definitive chemoradiotherapy. The study found that incidental irradiation of the sinoatrial node (SAN) during chemoradiotherapy may be associated with the development of AF and increased mortality. The maximum dose delivered to the SAN (SAN Dmax) exhibited the highest predictive value for AF development in both cohorts. Increased SAN Dmax was significantly associated with poor overall survival in both cohorts. The coronary artery calcium score was automatically determined using a software (AVIEW CAC, Coreline Soft) and was expressed as the Agatston score. The results support minimizing radiation dose exposure to the SAN during radiotherapy planning and considering close follow-up for the early detection of AF in patients receiving thoracic irradiation.

The study aimed to evaluate the reliability and agreement of a commercial deep learning-based software (AVIEW CAC, Coreline Soft) for fully automated coronary artery calcium (CAC) scoring on non-ECG-gated low-dose CT (LDCT) with different slice thicknesses compared with manual ECG-gated calcium-scoring CT (CSCT). The results showed that both CSCT and LDCT using the fully automated CAC-scoring software demonstrated excellent reliability and agreement with manual CSCT scoring. LDCT with a 1.0-mm slice thickness yielded more accurate Agatston scoring than LDCT with a 2.5-mm slice thickness using the fully automated commercial software.

In this study, researchers aimed to assess the feasibility and accuracy of a fully automated artificial intelligence (AI) powered coronary artery calcium scoring (CACS) method on ungated CT scans in oncologic patients undergoing 18F-FDG PET/CT. A total of 100 oncologic patients were retrospectively analyzed, comparing manual CACS on non-contrast ECG-gated CT scans with AI-CACS performed using AVIEW CAC (Coreline Soft) on ungated CT scans from 18F-FDG-PET/CT examinations. The AI-CACS tool demonstrated a sensitivity and specificity of 85% and 90% for detecting CAC, with an interscore agreement of 0.88 between manual and AI methods. Despite a generally underestimated CAC load on ungated CT by AI-CACS, the study concluded that fully automated AI-CACS using AVIEW CAC is feasible and provides an acceptable to good estimation of CAC burden in non-contrast free-breathing, ungated CT scans from 18F-FDG-PET/CT examinations.

This study aimed to investigate the potential and clinical utility of artificial intelligence (AI)–based computer-aided detection (CAD) of lung nodules in identifying pulmonary oligometastases. The chest CT scans of patients with lung metastasis from colorectal cancer were analyzed in two cohorts. In the first cohort, the CAD-assisted radiation oncologist (CAD-RO) demonstrated sensitivity and specificity for identifying oligometastatic disease (OMD) compared to an expert radiologist, with the sensitivity of the CAD-RO in nodule detection at 81.6%. The interobserver variability analysis showed an average sensitivity of 80.0%. In the second cohort, the 5-year survival rates were evaluated based on the number of CAD-RO–detected nodules, revealing a correlation between the number of nodules and survival. The study concluded that the use of AI, particularly employing the AVIEW LCS nodule detection AI, in OMD recognition demonstrates promising practicality, and the development of a deep learning–based model specific to the metastatic setting is underway.

This study aimed to validate coronary artery calcium (CAC) scores as prognostic factors for acute coronary events (ACE) development in breast cancer patients with cardiac exposure to radiation. Researchers evaluated breast cancer patients who received (n = 511) or did not receive (n = 600) adjuvant radiotherapy (RT) between 2005 and 2013. CAC Agatston scores were analyzed using the deep-learning-based algorithm, AVIEW-CAC (Coreline Soft). The study found that CAC scores were a significant predictor of ACE in patients with breast cancer. Over a 9.3-year follow-up period, the 10-year ACE rate was 0.7%. The results suggest that CAC score screening on simulation CT in patients undergoing breast RT can help identify those with a high risk for ACE on a per-patient basis.

Chronic obstructive pulmonary disease (COPD) is a complex and heterogeneous condition characterized by varying combinations of emphysema, small airways disease, and chronic bronchitis. In the past, quantification of chest CT imaging had limitations in its application to clinical practices. However, recent advances in pulmonary imaging technologies, particularly the use of artificial intelligence (AI), have significantly improved the quantification of chest CT imaging. In this article, the authors, who are researchers at Coreline Soft, discuss the current situation, limitations, and future role of chest CT imaging quantification in the assessment of COPD. They highlight the potential of AI-based solutions, such as those developed by Coreline Soft, to transform the diagnosis, evaluation, and management of this complex disease.

Chronic obstructive pulmonary disease (COPD) is a complex and heterogeneous condition characterized by varying combinations of emphysema, small airways disease, and chronic bronchitis. Previously, quantification of chest CT imaging had limitations in its application to clinical practices. However, recent advances in pulmonary imaging technologies, particularly the use of artificial intelligence (AI), have significantly improved the quantification of chest CT imaging. In this article, the authors discuss the current situation, limitations, and future role of chest CT imaging quantification in the assessment of COPD, showcasing numerous examples from AVIEW COPD by Coreline Soft. These AI-driven advancements hold the potential to transform the diagnosis, evaluation, and management of this complex disease.

The purpose of this study was to identify clinically relevant deep learning algorithms for emphysema quantification using low-dose chest computed tomography (LDCT) through an invitation-based competition. The Korean Society of Imaging Informatics in Medicine (KSIIM) organized this challenge, where seven research teams, including Coreline Soft, participated. A total of 558 CT scan pairs from nine hospitals were collected to develop algorithms that converted LDCT to simulate standard-dose CT (SDCT) values. The agreement between SDCT and LDCT was evaluated using various statistical measures. The results showed that emphysema quantification with LDCT was feasible through deep learning-based CT conversion strategies. Coreline Soft participated in the challenge and demonstrated strong performance, contributing to the development of effective AI solutions for emphysema quantification in real-world clinical settings.

The authors proposed a supervised multi-task aiding representation transfer learning network (SMART-Net) for intracranial hemorrhage (ICH) classification and segmentation using non-contrast head computed tomography (NCCT). The framework has upstream and downstream components, focusing on feature extraction and transfer learning for volume-level tasks. Experimental results from four test sets demonstrated that SMART-Net outperforms previous methods in robustness and performance for ICH classification and segmentation. The deep learning algorithms have been commercialized as part of the AVIEW NeuroCAD system by Coreline Soft, showing potential for improved emergency medical care.

This retrospective study investigated the value of thinner slice reconstructions in predicting major adverse cardiac events (MACEs) using cardiac CT. The study included 550 asymptomatic individuals and compared coronary artery calcium (CAC) scores from 1.5 mm and 3 mm scans. CAC scores from 1.5 mm scans were significantly higher, but no difference was found in the predictive accuracy of MACEs between the two scan types. In conclusion, CAC scores from 1.5 mm scans do not provide added prognostic value compared to 3 mm scans. The CAC scores were evaluated using commercial software, AVIEW CAC by Coreline Soft.

This study aimed to describe quantitative CT parameters in COVID-19 patients based on disease severity and build decision trees for predicting respiratory outcomes using these parameters. Patients were classified into four categories based on disease severity. High attenuation area (HAA) was defined and showed an increase with severity. Whole-lung images were analyzed using commercial software Aview system by Coreline Soft Inc. Decision tree models were built using clinical variables and initial laboratory values, with and without quantitative CT parameters. Results showed that HAA had a moderate correlation with lactate dehydrogenases (LDH) and C-reactive protein (CRP). The decision tree model including quantitative CT parameters showed greater accuracy in predicting respiratory failure. In conclusion, using quantitative CT parameters along with clinical characteristics, PCR Ct value, and blood biomarkers can provide higher accuracy in predicting respiratory failure.

COPD is a heterogenous condition, and accurate classification can help tailor therapies and improve patient outcomes. Quantitative computed tomography (QCT) techniques show promise in assessing COPD phenotypes. Companies have commercialized automated CT software packages, like Coreline Soft AVIEW, which are increasingly used in research and clinical settings. This article reviews key features of such platforms, their underlying technologies, and clinical applications. However, their full potential in diagnostic work-up and treatment planning remains to be explored.

This position paper presents recommendations for low dose computed tomography (LDCT) scan protocols and reading in lung cancer screening (LCS), as LDCT has been proven to decrease lung cancer mortality in smokers and former smokers. Several Italian initiatives are offering LCS and smoking cessation to high-risk subjects. The paper details fundamentals for classifying lung nodules and other findings at LDCT tests, along with international guidelines for their reporting and management in LCS. The Italian College of Thoracic Radiologists aims to establish a homogenous, evidence-based approach to LDCT tests in Italy. Although the software is not directly mentioned, screen captures from AVIEW COPD were used in the article.

This study aimed to assess the accuracy of a fully automated deep learning (DL) based coronary artery calcium scoring (CACS) tool using non-contrast CT scans acquired for attenuation correction (AC) of cardiac SPECT-MPI. A total of 56 patients with suspected coronary artery disease were prospectively enrolled. CACS was manually assessed and compared to the cloud-based DL tool's results. The interscore agreement between the standard of reference and the DL tool was 0.986, with a 0.977 agreement in risk categories and a 3.6% reclassification rate. Factors such as heart rate, image noise, BMI, and scan did not significantly impact the absolute percentage difference in CAC scores. The study concluded that the DL tool (AVIEW CAC, Coreline Soft) enables accurate and fully automated estimation of CAC scores in patients undergoing non-contrast CT for AC of SPECT-MPI.

This study investigated the correlation between longitudinal changes in pulmonary vascular parameters and clinical parameters in 288 COPD patients over a follow-up period of up to 6 years. CT images were classified into five subtypes based on the percentage of low-attenuation areas. The total number of vessels and the total number of vessels with an area less than 5 mm2 per 1 cm2 of lung surface area (LSA) were measured. Results showed that these vascular parameters changed during the progression of GOLD stages and were negatively correlated with LAA-950 and LAA-856exp. Pulmonary function test (PFT) results showed a weak correlation with these parameters. Quantitative CT features faithfully reflected pulmonary vessel alterations, demonstrating that pulmonary vascular alteration started in COPD patients. The vessel analysis function in AVIEW COPD by Coreline Soft was utilized in this study.

The deep learning-based software, AVIEW CAC from CORELINE Soft, demonstrated excellent performance in a population-based screening setting for risk categorization in asymptomatic participants. The study suggested that future deep learning software with the ability to assign uncertain cases for manual human feedback could improve calcium scoring and outperform experienced readers using manual scoring alone. However, the study's limitations include its focus on a cardiac screening setting with a highly standardized protocol, which may limit the generalizability of the results to clinical settings with different populations and scanner settings. In conclusion, the deep learning-based software can accurately categorize risk and initiate preventive treatment in a cardiovascular CT screening setting.

The study evaluated the performance of an AI-based lung cancer screening prototype, AVIEW LCS from Coreline Soft, compared to experienced radiologists in ultra-low-dose CT scans for lung cancer baseline screening. 283 participants were included, and the AI was able to automatically detect, measure, and classify solid nodules. Results showed that the AI outperformed radiologists in negative-misclassifications, indicating that AI could reduce radiologists' workload by up to 86.7%. The use of AI in lung cancer screening could also improve standardization and implementation. However, the study had limitations as it was focused on a specific screening setting with a highly standardized protocol, and its generalizability to other settings may be limited.

This study aimed to investigate differences in femur geometry between patients with atypical femur fractures (AFFs) and the general population and evaluate biomechanical factors related to femoral bowing in AFFs. Retrospective data from 46 patients were evaluated, including femur computed tomography images that were reconstructed into 3D models analyzed with geometry analysis software to obtain femur length, width, and curvature. Patients were divided into subtrochanteric and shaft AFF groups and compared to a general population of 300 women. The femur CT images were imported into AVIEW Modeler, a 3D modeling software from Coreline Soft, to produce the 3D samplings. The study found that femoral bowing and width increased in shaft AFFs but were similar to subtrochanteric AFFs compared to the general population.

The article provides a comprehensive overview of the current status of lung cancer in Korea, covering epidemiology, screening, diagnosis, and treatment. It explains also the K-LUCAS program, a lung cancer screening program in Korea that has implemented a cloud-based system using computer-aided detection and volumetry software from Coreline Soft. The use of this software has led to an increase in positive cancer findings, but has also reduced variability among screening centers. Overall, the article emphasizes the importance of effective lung cancer screening programs and the potential benefits of using advanced software tools to improve the accuracy and consistency of screening results.

his study aimed to investigate whether the psoas muscle volume measured from pelvic CT scans could be used as a diagnostic tool to assess muscle mass and sarcopenia in patients with hip fractures. The study included 57 patients who underwent surgery for hip fractures, and their psoas muscle volume was calculated using a 3D modeling software program (AVIEW Modeler, Coreline Soft, Seoul, Korea). The results showed that psoas muscle volume had a stronger correlation with appendicular skeletal muscle mass and handgrip strength compared to the cross-sectional area of the psoas muscle. The psoas muscle volume could be used as a potential diagnostic tool to assess the quantity of skeletal muscle in patients with hip fractures without an additional examination.

The study aimed to develop an automated method for imaging and tracking the position of the inferior alveolar nerve (IAN) using artificial intelligence (AI) in cone-beam computed tomography datasets. A customized 3D nnU-Net was used for image segmentation, and active learning was carried out in iterations for 83 datasets. The accuracy of the model for IAN segmentation was evaluated using the remaining 50 datasets, and the dice similarity coefficient (DSC) value and the segmentation time for each learning step were compared. The deep active learning framework was found to be a fast, accurate, and robust clinical tool for demarcating IAN location. The overall segmentation was performed using AVIEW Modeler software (version 1.0.3, Coreline Software, Seoul, Korea), and the ground truth of the IAN was provided by three specialists.

The CAC_auto system, which is a deep learning-based fully automatic calcium scoring system, was validated in this study using previously published cardiac CT cohort data, with the manually segmented coronary calcium scoring (CAC_hand) system as the reference standard. The AVIEW CAC software was used for the automatic CAC measurement. The results showed that the CAC_auto system provided accurate calcium score measurement and risk category classification, with high sensitivity and low false-positive rates in detecting coronary calcium, as well as high reliability in measuring the Agatston score per vessel and per patient. The main causes of false-positive results were image noise, aortic wall calcification, and pericardial calcification. The study suggests that the CAC_auto system could potentially streamline CAC imaging workflows.

This retrospective study evaluates the impact of a content-based image retrieval (CBIR) system utilizing deep learning on the diagnostic accuracy of interstitial lung disease (ILD) in chest CT images. The system, employing Coreline Soft's AVIEW Lung Texture algorithm, was tested on 288 patients with confirmed ILD and four different disease classes. The diagnostic process involved eight readers of varying experience levels, and results were compared before and after the application of CBIR. Findings revealed a significant improvement in overall diagnostic accuracy (from 46.1% to 60.9%), particularly in cases of usual interstitial pneumonia (UIP) and nonspecific interstitial pneumonia (NSIP). Moreover, interreader agreement also improved markedly. The study concludes that the deep learning-enhanced CBIR system notably enhances diagnostic accuracy and interreader consensus in ILD cases, highlighting the potential utility of AVIEW software for future commercialization.

The study evaluates the performance of Coreline Soft's AVIEW RT-ACS, a deep learning-based autocontouring system, in delineating organs at risk (OARs) during breast radiotherapy. Eleven experts were initially tasked with manually contouring nine OARs in 10 cases of adjuvant radiotherapy post breast-conserving surgery, which were then corrected using the autocontouring system. Comparisons between manual contours, corrected autocontours, and autocontours were made using Dice similarity coefficient (DSC) and Hausdorff distance (HD). The results demonstrated that the autocontouring system's accuracy was comparable to that of the experts' manual contouring, with an average DSC value of 0.90. Furthermore, the system significantly reduced interphysician variations and decreased contouring time from 37 minutes (manual) to 6 minutes (corrected autocontours). The AVIEW RT-ACS system demonstrated good user satisfaction, highlighting its potential to enhance the quality of breast radiotherapy and reduce variability in clinical practice.

This study explores the potential of a novel quantitative CT emphysema air-trapping composite (EAtC) mapping method, implemented in Coreline Soft's AVIEW COPD, for assessing lung function in COPD patients. The method classifies lung parenchyma into three areas - normal, functional air trapping (fAT), and emphysema - based on the analysis of inspiration and expiration CTs in 584 COPD patients. The volume fractions of each area were then compared with clinical parameters and pulmonary function tests (PFTs). Findings revealed that each EAtC class had moderate correlations with forced expiratory volume in 1 second (FEV1) and FEV1/forced vital capacity (FVC). Both fAT and emphysema were significant predictors of FEV1 and FEV1/FVC, with fAT showing better correlations with PFTs than small-airway disease on parametric response mapping. The study concludes that EAtC mapping offers comprehensive lung functional information, potentially serving as an imaging biomarker of lung function in COPD.

This study presents the development and validation of a deep learning-based automatic cardiovascular border (CB) analysis algorithm (CB_auto) for diagnosing and quantitatively evaluating valvular heart disease (VHD) using posterior-anterior chest radiographs (CXR). The CB_auto, developed using 1614 normal and VHD CXRs, demonstrated excellent reliability with an intraclass correlation coefficient above 0.98. When compared to manual CB drawing (CB_hand), the absolute percentage measurement error was under 10% for all parameters except for carinal angle and left atrial appendage. In addition, CB parameters were found to be significantly greater in VHD than in normal controls, and correlated significantly with echocardiographic measurements. This suggests that CB_auto could serve as a reliable tool for daily clinical practice and research purposes, enhancing diagnosis and evaluation of VHD. The proposed method is set to feature in a forthcoming product from Coreline Soft.

This research applies radiomics analysis to predict overall survival in patients with chronic obstructive pulmonary disease (COPD). The study, which includes 344 patients from the Korean Obstructive Lung Disease cohort and an external validation cohort of 112 patients, identifies five features through the least absolute shrinkage and selection operation (LASSO) Cox regression analysis. These features form a radiomics signature (RS), which effectively classifies COPD patients into high or low mortality risk groups. The RS demonstrated a good performance, with a C-index of 0.774 in the discovery group and 0.805 in the validation group. The high-risk group identified by the RS exhibited worse overall survival. Overall, the study confirms the feasibility of using a radiomics approach for survival prediction and risk stratification in COPD patients, with the radiomics model showing promising performance. This highlights the feasibility of a radiomics approach for survival prediction and risk stratification in COPD patients, and the potential utility of the AVIEW COPD in this regard.

This study aims to segment the maxillary sinus into maxillary bone, air, and lesion, by employing an active learning framework in conjunction with a customized 3D nnU-Net. Tested on 83 randomly selected cases, the framework uses limited cone-beam computed tomography (CBCT) datasets, reducing annotation efforts and costs. The approach shows consistent accuracy for air and lesion segmentation across three progressive stages, as indicated by the Dice similarity coefficients (DSCs). The use of this framework also significantly decreases the time required for segmentation. Overall, the findings demonstrate that the deep active learning framework can efficiently train on limited CBCT datasets, potentially enhancing segmentation tasks in dental CBCT images. This enhancement is particularly noticeable when using Coreline Soft's AVIEW Modeler software.

This study evaluated the relationship between visually assessed computed tomography (CT) subtypes of Chronic obstructive pulmonary disease (COPD) and various clinical and imaging characteristics. Between 2012 and 2017, 452 participants were enrolled and seven COPD subtypes were defined using visual evaluation of CT images. The study found significant differences in clinical, laboratory, spirometric, and quantitative CT features among these CT-based visual subtypes, the latter of which was analyzed using the Aview system by Coreline Soft. As the visual grade of centrilobular emphysema worsened, there was a noted decline in pulmonary function and an increase in COPD assessment test score and quantitative assessment. Conversely, the bronchial subtype was associated with a higher body mass index, better lung function, and higher lung density. The study concluded that classifying COPD subtypes using visual CT imaging features can reflect the disease's heterogeneity and pathological processes.

This study aimed to identify the optimal diameter threshold for defining positive results in solid nodules at baseline screening low-dose CT (LDCT) and to compare the effectiveness of two-dimensional and volumetric measurements of lung nodules in diagnosing lung cancers. The study population consisted of consecutive participants from the Korean Lung Cancer Screening project between 2017 and 2018. Semi-automated segmentation was used to measure the average transverse diameter and effective diameter of lung nodules using the A-view Lungscreen software from Coreline Soft. The diagnostic performances for lung cancers, diagnosed within 1 year after LDCT, were evaluated using area under receiver-operating characteristic curves (AUCs), sensitivities, and specificities. The study found that elevating the diameter threshold for solid nodules from 6 to 9 mm could significantly increase the specificity and modestly reduce sensitivity, resulting in a 60.2% estimated reduction of unnecessary follow-up LDCTs, and a diagnostic delay in 1.9% of lung cancers. Thus, the results suggest that increasing the diameter threshold for solid nodules may lead to a substantial decrease in unnecessary follow-up LDCTs with a small proportion of diagnostic delay of lung cancers.

This study, known as the Korean Lung Cancer Screening Project (K-LUCAS), was a single-arm cohort study conducted from February 2017 with the aim to evaluate the feasibility of implementing a national lung cancer screening program in Korea. The study targeted a high-risk population aged 55–74 years with more than a 30-pack-year smoking history. The results of the screening were reported using the Lung Imaging Reporting and Data System (Lung-RADS), as recommended by the American College of Radiology. To maintain the quality of the screening, K-LUCAS used a network-based diagnostic supporting system with a computer-aided detection (CAD) program, specifically AVIEW LCS nodule detection AI developed by Coreline Soft. (This software was not directly mentioned in the paper but was used for the process.) The study found that among the 5,692 eligible high-risk participants, 865 had positive screening results requiring further examination and 42 had confirmed lung cancer. The results of K-LUCAS provide encouraging evidence supporting the implementation of a national lung cancer screening program to detect early stage lung cancer and promote smoking cessation among participants in the Asian population.

This study compared manual and deep learning-based auto-segmentation (AVIEW RT-ACS, Coreline Soft) for delineating target volumes and organs-at-risk (OARs) in breast cancer patients receiving radiotherapy. The algorithm demonstrated a strong correlation with manual contours, achieving a Dice similarity coefficient of over 0.80 for all OARs and over 0.70 for all breast and regional lymph node clinical target volumes. The study also found minor differences in dosimetric parameters between the auto-segmented and manual contours. While not replacing radiation oncologists, deep learning-based auto-segmentation proved to be a potential tool to assist them, potentially reducing workload and inter-physician variability. The study concludes that auto-segmentation using AVIEW RT-ACS is feasible and beneficial in breast radiotherapy planning.

This study evaluated the feasibility of using AVIEW RT-ACS, a deep learning-based auto-segmentation software developed by Coreline Soft, for target and organs-at-risk (OAR) segmentation in breast cancer patients undergoing radiotherapy (RT). The software was used to auto-segment clinical target volumes (CTV) and OARs of 61 patients who underwent breast-conserving surgery. The study found good correlation between the auto-segmented and manual contours for most OARs and CTVs, except for the right coronary artery and left anterior descending artery. The study suggests that auto-segmentation can be an expedient tool to assist radiation oncologists and improve the quality control of RT in breast cancer patients.

In this study, the performance of a content-based image retrieval (CBIR) system for diffuse interstitial lung disease (DILD) was evaluated. The CBIR system utilized a convolutional neural network to automatically quantify and classify six image patterns of DILD from chest CT scans. A total of 246 pairs of initial and follow-up chest CT scans from patients with usual interstitial pneumonia, nonspecific interstitial pneumonia, and cryptogenic organic pneumonia were used to assess the CBIR system's performance. Sixty cases were selected as queries, and the CBIR system retrieved five similar CT scans from the database. The results showed that the CBIR system had a high rate of retrieving the same pairs of query CT scans in the top 1-5 retrievals and could retrieve similar CT scans more accurately in UIP compared to NSIP and COP. Radiologists also rated a high percentage of retrieved CT scans as having a high degree of similarity to the query CT scans. The study demonstrates the potential utility of the CBIR system for improving the diagnosis and management of DILD, and highlights the potential for commercialization of AVIEW software.

This study evaluated the variability of computer-assisted interpretation of low-dose chest CTs among radiologists in a nationwide lung cancer screening program, using a dedicated reading system (AVIEW Lungscreen, Coreline Soft). The results showed substantial inter-institution variability in interpreting screening results, attributed to different usage of the computer-assisted system. The study included 3353 baseline low-dose chest CTs, initially interpreted by 20 radiologists from 14 institutions, and then re-interpreted by a single radiologist in a retrospective central review. Positive rates by Lung-RADS varied from 7.5% to 43.3% across radiologists, while rates by the NELSON criteria varied from 11.4% to 45.0%. The central review demonstrated higher positive rates and lower inter-institution variability compared to the initial institutional reading, indicating a significant impact of individual radiologist's discretion on the interpretation of lung cancer screening results.

The study aimed to evaluate the impact of a cloud-based CT interpretation system equipped with semi-automated measurement and computer-aided detection (CAD) for lung nodules on a nationwide lung cancer screening program. The program initially relied on each institution's interpretation systems and manual nodule measurement. The cloud-based system, implemented by Coreline Soft using AVIEW LCS, was later introduced. Results showed that while the number of detected nodules significantly increased with the cloud-based system, the positive rate didn't significantly differ from the conventional system. However, the variability of positive rates across institutions significantly reduced with the cloud-based system. In terms of sensitivity and specificity based on Lung-RADS version 1.0, no significant difference was observed between the two systems. The findings suggest that CAD and semi-automated measurements can enhance nodule detection and reduce variability in positive rates across different institutions in a nationwide lung cancer screening program.

The study aimed to map the course of the radial nerve at the posterior humeral shaft using a three-dimensional (3D) reconstruction technique based on computed tomography (CT) images of living subjects. This was to prevent iatrogenic radial nerve injury, a significant complication during surgical exposure of the humerus. The in vivo study utilized 652 upper extremity CT angiography images from 326 patients. The 3D modeling of the humerus and radial nerve was performed using AVIEW Modeler 1.0 by Coreline Soft Inc., a software capable of automatic segmentation, 3D reconstruction, and measurements. The proximal, mid, and distal points of the radial nerve on the posterior aspect of the humeral shaft were measured. The results demonstrated that the 3D reconstruction technique provides a reliable map of the radial nerve course and these measurements showed excellent intraobserver and interobserver reliability.

The study aimed to evaluate the reproducibility and postoperative stability of a 3D printed surgical guide in mandibular reconstruction using osteocutaneous free flaps. Fifteen patients underwent simulation surgery with 3D surgical guides fabricated using preoperative CT images (T0) on 3D modeling software platforms Aview(R) Modeler by Coreline Soft or OnDemand3D by Cybermed. Postoperative CT scans were taken immediately, one week (T1), and six months (T2) after surgery. The study found the error of the 3D surgical guide ranged from 0.85 to 2.56 mm, with no differences in reproducibility based on the type of flap used. However, condylar error and error at the mandible midpoint were significantly different in fibular free flap (FFF). Six months post-surgery, condyles moved more than 2 mm in FFF but showed no significant movement in deep circumflex iliac artery free flap (DCIA). The findings suggest careful intraoperative flap fixation and close postoperative observation, especially for FFF cases.

The study aimed to evaluate the efficiency and accuracy of an auto-segmentation framework using a 3D-patch-based convolutional neural network, U-Net, enhanced with a graph-cut algorithm, in segmenting abdominal organs in radiotherapy. The framework utilized 3D-patch-based CT images for the liver, stomach, duodenum, and right/left kidneys. The segmentation accuracy was assessed by comparing the predicted structures with those produced from atlas-based methods and inter-observer segmentation, using the Dice similarity coefficient, Hausdorff distance, and mean surface distance. The U-Net-based auto-segmentation outperformed atlas-based methods for all abdominal structures and showed comparable results to inter-observer segmentation, particularly for the liver and kidneys. Moreover, the average segmentation time reduced from 22.6 minutes to 7.1 minutes with automation using U-Net, showcasing potential clinical usefulness in terms of accuracy and time-efficiency. The concept of this study has been incorporated into AVIEW RT-ACS by Coreline Soft.

This study aimed to identify the optimal method for evaluating coronary artery calcium (CAC) severity on non-electrocardiogram-gated low-dose chest computed tomography (LDCT) within the Korean Lung Cancer Screening (K-LUCAS) project. Four cardiothoracic radiologists independently assessed CAC severity using four different scoring methods and classified severity for each case as none, mild, moderate, or severe. The interobserver agreement was found to be moderate for visual assessment and substantial for the other three methods. Agreement between the four grading methods for the same observer was substantial for three observers and moderate for one. The mean evaluation time varied significantly between methods. The study concluded that artery-based grading could be appropriate due to its substantial interobserver agreement and acceptable mean evaluation time. The K-LUCAS project has been managed by Coreline Soft Inc and the study used the AVIEW Research image review platform.

This secondary analysis of the Korean Lung Cancer Screening Project (K-LUCAS) evaluated the impact of tuberculosis sequelae on CT lung cancer screening effectiveness. Out of 11,394 high-risk participants, 1,509 (13%) had tuberculosis sequelae. The study found that while these sequelae were linked to positive CT screening results, they didn't correlate with lung cancer diagnosis. It also revealed that the Lung-RADS system's specificity was lower in participants with tuberculosis sequelae compared to those without. The study concluded that tuberculosis sequelae could potentially reduce the specificity of Lung-RADS-based CT screening for lung cancer in at-risk populations. During the project, the A-view Lungscreen system by Coreline Soft was implemented.

This study investigated variability in the interpretation of nationwide lung cancer screening CT scans, focusing on the potential of computer-aided detection and volumetry in reducing interobserver variability. The analysis involved 1,647 baseline screening CT scans from a Korean nationwide lung cancer screening project (K-LUCAS). The scans were interpreted using the AVIEW Lungscreen system implemented by Coreline Soft. There was a significant difference in per-case positive rates and per-nodule positive rates between institutional readings and a central review. The central review detected more nodules per case, but these nodules were smaller. Manual modification of semi-automated segmentation was more common in institutional readings. Inter-institution variability in positive rates was lower in the central review. The study concluded that despite the use of computerized tools, significant variability exists in CT interpretations for lung cancer screening, largely due to discarding tiny nodules and modifying segmentation results. Further investigation is needed to ascertain the benefits of reducing variability by applying stricter rules.

A patient with squamous cell carcinoma of the maxillary sinus underwent a unilateral total maxillectomy and anterolateral thigh (ALT) free flap reconstruction. Post-surgery, a large maxillary defect without a bone graft led to midfacial collapse, exacerbated by radiotherapy. Given the aesthetic importance of 3-dimensional positioning in maxillary reconstruction, a mirror image was created using the normal contralateral side. The virtually reconstructed maxilla was evaluated for function and symmetry through computer simulation, facilitated by AVIEW Modeler. A surgical guide was created using a 3D printing system. Guided by this, a deep circumflex iliac artery (DCIA) free flap was harvested and implemented in the reconstruction, resulting in satisfactory outcomes. The usage of mirror image-based virtual surgical planning and 3D printed guides, as enabled by AVIEW Modeler, proved highly effective in maxilla reconstruction with DCIA flaps.

We developed a robust lung segmentation method using a deep convolutional neural network (CNN) for high-resolution computed tomography (HRCT) and volumetric CT images of diffuse interstitial lung disease (DILD). Using scans of 617 patients with DILD types such as cryptogenic organizing pneumonia, usual interstitial pneumonia, and nonspecific interstitial pneumonia, we created gold standards with manually corrected segmentation by an expert thoracic radiologist. We then used a two-dimensional U-Net architecture with deep CNN for lung region segmentation in HRCT images. The U-Net-based segmentation method was found to be highly accurate, showing significant improvement over conventional methods (p < 0.001) for both HRCT and volumetric CT. The insights and methods of this study have been comercialized in Coreline's AVIEW platform, thereby promising its effective use for different clinical protocols.

This study aimed to assess the accuracy of computed tomography (CT) texture analysis (CTTA) in differentiating benign from malignant mediastinal lymph nodes (LNs) in non-small-cell lung cancer patients. We retrospectively reviewed pathologically confirmed malignant and benign mediastinal LN samples, obtained via endobronchial ultrasound-guided transbronchial needle aspiration, alongside chest CT and 18-fluorodeoxyglucose (FDG) uptake positron emission tomography data. CTTA was performed using the "AVIEW" software by Coreline Soft. Among the 132 LNs from 80 patients analyzed, compactness and normalized standard deviation in CTTA features showed differences between malignant and benign LNs. The ability to distinguish malignant LNs was higher using high standard uptake value on FDG PET/CT and normalized SD on CTTA; however, normalized SD had low sensitivity despite high specificity. Thus, CTTA may aid in distinguishing between benign and malignant LNs, but its diagnostic value was not high, necessitating integrated evaluation with other imaging modalities.

This study developed and validated a fully automated lung lobe segmentation method with 3D U-Net, using chest CT scans of 196 normal and mild-to-moderate COPD patients from three centers. Manual correction by a thoracic radiologist served as the gold standard for comparison. The deep learning method showed high accuracy in both internal and external validation when compared to the gold standards using metrics such as Dice similarity coefficient (DSC), Jaccard similarity coefficient (JSC), mean surface distance (MSD), and Hausdorff surface distance (HSD). The segmentation method was also time-efficient. Despite challenges in developing a robust automatic lung lobe segmentation method, this deep learning–based 3D U-Net method demonstrated promising accuracy and computational time, and could potentially be adapted for severe lung diseases in clinical workflows. The algorithm used for this method has been commercialized in AVIEW by Coreline Soft.

This study aimed to assess correlations between computed tomography (CT) parameters and pulmonary function test (PFT) parameters in 370 patients with chronic obstructive pulmonary disease (COPD). Patients were classified according to disease severity using the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria. Emphysema index (EI), air-trapping index, and airway parameters were measured using the automatic segmentation software, Aview, developed by Coreline Soft, Seoul, Korea. The findings indicated that pulmonary function deteriorated as disease severity increased. Parenchymal attenuation parameters worsened in patients with higher GOLD stages, while airway parameters showed a mild correlation with PFT results. For patients with mild COPD, both airway and parenchymal attenuation parameters independently predicted pulmonary function. However, in patients with severe COPD, parenchymal attenuation parameters were the dominant predictors of pulmonary function. These results highlight the potential utility of CT parameters in predicting PFT indices in COPD patients.

We present a novel airway segmentation method for volumetric chest computed tomography (CT) scans and assess its performance on multiple datasets. Our method employs a 2.5D convolutional neural network (CNN) trained in a supervised manner to segment the airway voxel-by-voxel. To enhance accuracy, we utilize three adjacent slices in orthogonal directions (axial, sagittal, and coronal) and fine-tune parameters related to tree length and leakage. The gold standard for airway segmentation was generated using a semi-automated method in AVIEW. We trained and evaluated the 2.5D CNN on inspiratory thoracic CT scans from the Korean obstructive lung disease study, including subjects with normal lung function and those with chronic obstructive pulmonary disease (COPD). The reliability and practicality of our method were demonstrated across multiple datasets. In eight test datasets following the same imaging protocol, our method achieved a 92.16% detection rate, 7.74% false positive rate, and a Dice similarity coefficient of 0.8997 ± 0.0892. In 20 test datasets from the EXACT'09 challenge, our method achieved a 60.1% detection rate and a 4.56% false positive rate.

This study aimed to identify predictive factors for treatment responsiveness in 226 patients with chronic obstructive pulmonary disease (COPD) using baseline CT scans. Patients received a combination of inhaled long-acting beta-agonists and corticosteroids. Quantitative analysis of the scans revealed that the IBHB-measured Pi10 parameter was the only independent variable significantly associated with an increase in forced expiratory volume in 1 second (FEV1) after treatment (p = 0.003, adjusted odds ratio: 1.787). The researchers utilized the commercial software AVIEW from Coreline Soft to accurately measure standardized airway dimensions on the CT scans, enabling the prediction of treatment responsiveness in COPD patients. This finding contributes to the development of personalized treatment strategies based on quantitative CT scan analysis.

This study aimed to automatically extract pulmonary vessels from volumetric chest CT scans and evaluate their associations with clinical parameters, including survival, in patients with chronic obstructive pulmonary disease (COPD). Pulmonary vessels were successfully extracted using the AVIEW COPD lung vessel analysis software in 344 adult COPD patients. Quantitative pulmonary vessel features obtained from different lung surface areas (LSAs) showed significant associations with survival and pulmonary function tests. Specifically, the percentage of total vessel area measured at 18 mm depth from the pleural surface (%VA18mm) and the number of small vessels per 10 cm2 of LSA at 9 mm depth from the pleural surface (N<5mm/LSA9mm) demonstrated the strongest predictability for clinical parameters. Furthermore, significant correlations were observed between quantitative pulmonary vessel features and the extent of emphysema. The study highlights the technical feasibility and clinical relevance of automated pulmonary vessel extraction and quantitative assessment using the AVIEW COPD lung vessel analysis software, providing valuable insights into the relationship between pulmonary vascular alterations and COPD.

This study aimed to evaluate the accuracy of emphysema volume (EV) and airway measurements (AMs) using different iterative reconstruction (IR) algorithms and virtual monoenergetic images (VME) at low- and standard-dose settings. Computed tomography (CT) scans were obtained on a phantom and reconstructed using various algorithms. EV and airway wall area percentage (WA%) were measured, and absolute percentage measurement errors (APEvol and APEWA%) were calculated. Results showed that IMR-R1 provided the most accurate measurement of EV, while IMR-SP1 was the most accurate for AM in both low- and standard-dose CT. No significant differences were found in APEvol between low- and standard-dose scans for IMR-R1. However, VME at 70 keV showed higher APEvol compared to other algorithms and higher APEWA% compared to all other algorithms. The study concluded that IMR was the most accurate technique for EV and airway measurements. The measurements were performed using the commercial software AVIEW from Coreline Soft, where automatic segmentation and volume quantification were carried out by clicking on each air column.

This study assessed a vessel removal algorithm's effectiveness in segmenting subsolid nodules by comparing software measurements of the solid component on CT scans, before and after vessel removal, with invasive component measurements on pathology in lung adenocarcinomas. Seventy-three nodules with ≤10 mm invasive components on pathology were analyzed. Semi-automated segmentation was performed by two radiologists. Software-derived 3D longest, axial longest, and effective diameters of the solid component were obtained pre- and post-vessel removal. Measurements were compared to pathology using statistical tests. Results revealed a significant correlation between software and pathology measurements, with improved correlation after vessel removal. Measurement differences between CT and pathology were significantly reduced post-vessel removal. The smallest difference was observed with the 3D longest diameter after vessel removal, showing no significant difference from pathology. Incorporating the vessel removal algorithm in software segmentation enhanced the prediction of the invasive component in lung adenocarcinomas. AVIEW Lung Screen software from Coreline Soft was used for segmentation.

A novel ground-glass nodule (GGN) segmentation method that separates solid components and ground-glass opacity (GGO) in chest CT images has been developed. This technique first extracts the initial solid component and GGO using intensity-based segmentation with histogram modeling. The extracted regions are then refined using an asymmetric multi-phase deformable model with an intensity-constrained averaging function. Finally, vessel-like structures are removed based on multi-scale shape analysis. Experimental evaluations using datasets from Seoul National University Hospital (SNUH) and Lung Image Database Consortium and Image Database Resource Initiative (LIDC/IDRI) demonstrated high segmentation accuracy for the proposed method, showing significant correlation with manual segmentation. This innovative approach enhances GGN segmentation accuracy, proving the effectiveness of the asymmetric multiphase deformable model and pulmonary vessel removal in chest CT image analysis. The proposed GGN segmentation algorithm have been transferred to AVIEW Lung Screen, Coreline Soft.

In this comprehensive review article, the authors explore the technical advancements in quantitative CT imaging for assessing Chronic Obstructive Pulmonary Disease (COPD). They highlight the potential of quantitative imaging parameters in providing precise and objective assessments of COPD-related components, such as emphysema, small-airway disease, and comorbidities like pulmonary vessel alterations, atherosclerosis, cachexia, and osteoporosis. The authors discuss the association between these parameters and clinical measures of disease severity, respiratory symptoms, COPD exacerbations, and mortality, showcasing promising findings from various studies. Despite the optimistic results, the article also addresses the challenges hindering the integration of quantitative CT assessment into everyday COPD management. While the authors don't directly mention AVIEW software, they include numerous example screenshots throughout the article, indirectly illustrating the application and potential benefits of quantitative CT imaging in assessing COPD. The review provides insights into the current status, clinical implications, limitations, and future prospects of quantitative CT imaging in COPD assessment.

This study presents a computer-aided differential diagnosis (CADD) system for differentiating usual interstitial pneumonia (UIP) and non-specific interstitial pneumonia (NSIP) using high-resolution computed tomography (HRCT) images. Six local interstitial lung disease patterns were identified, and a support vector machine (SVM) classifier was employed to label regions of interest based on texture and shape characteristics. The developed system's performance was then tested on 54 clinically diagnosed UIP and NSIP patients' HRCT images. Impressively, the CADD system achieved an accuracy of up to 91% after sequential-forward feature selection, outperforming thoracic radiologists whose accuracies ranged from 75% to 87%. The findings from this study later were transferred to the development of AVIEW Lung Texture by Coreline Soft, embodying these advanced diagnostic capabilities into commercial software to potentially revolutionize diagnostic radiology.

This retrospective study investigated whether the baseline extent and 1-year change in regional disease patterns on CT can predict the survival of patients with idiopathic pulmonary fibrosis (IPF). The research incorporated 144 IPF patients, utilizing an in-house texture-based automated system to quantify the extents of five regional disease patterns. The fibrosis score was defined as the sum of the extent of honeycombing and reticular opacity. Through the Cox proportional hazard model, the study determined that age, desaturation, fibrosis score, and interval change in fibrosis score were significant independent predictors of survival. This texture-based, automated CT quantification of fibrosis, which later influenced the development of AVIEW Lung Texture by Coreline Soft, can indeed serve as an independent predictor of survival in IPF patients, offering a valuable tool for predicting disease outcomes.

This study introduces a novel approach to size-based emphysema clustering, evaluating the size variation and collapse of holes in emphysema clusters using inspiratory and expiratory computed tomography (CT). The researchers developed a new method for the size differentiation of emphysema holes, which was applied to the volumetric CT results of emphysema patients. The study found significant correlations between blind visual evaluations and the classification using the size-based emphysema clustering method. Furthermore, the collapse of emphysema holes was found to be significantly correlated with pulmonary function test parameters. The study concluded that a detailed assessment of the size variation and collapse of emphysema holes could provide useful insight into the dynamic collapse of emphysema and its functional relationship. The methodology developed in this research has been commercialized in AVIEW COPD by Coreline Soft, offering a new tool for understanding and managing emphysema.

This study proposes the use of ensemble classifiers to address inter-scanner variations in differentiating regional disease patterns in high-resolution computed tomography (HRCT) images of diffuse interstitial lung disease patients from different scanners. The experiment involved 600 regions of interest (ROIs) and 92 HRCT images from two different scanners, classified by expert radiologists. Textual and shape features were extracted from the ROIs and the whole lung parenchyma. Ensemble classifiers showed higher accuracy compared to individual classifiers in the integrated scanner test, as well as partial improvements in intra- and inter-scanner tests. In the whole lung classification, the ensemble classifiers achieved higher quantification accuracies compared to individual classifiers. Overall, the study concludes that ensemble classifiers perform better, particularly with integrated scanner images, providing a promising approach to address inter-scanner variations in HRCT analysis. The algorithm has been comercialized in AVIEW Lung Texture, Coreline Soft.

In this retrospective study, the objective was to identify predictive factors for a decline in forced vital capacity (FVC) based on initial CT scans using texture-based automated quantification in patients with idiopathic pulmonary fibrosis (IPF). The study included 193 IPF patients with 1-year follow-up pulmonary function tests. A texture-based automated system was used to quantify six regional CT patterns: normal, ground-glass opacity (GGO), reticular opacity (RO), honeycombing, emphysema, and consolidation. A decline in FVC was defined as a decrease of more than 10% in the initial FVC. The results showed that a decline in FVC occurred in 32 patients, with the extent of RO being the only independent predictor for this decline. A receiver operating characteristic (ROC) analysis indicated that an RO extent of less than 22.05% accurately predicted stable IPF at the 1-year follow-up in terms of FVC. The concept of the research has been comercialized in AVIEW Lung Texture, Coreline Soft.

This study aimed to determine the power-law exponents (D) of emphysema hole-size distributions as a reliable index and investigate their robustness to breath-hold-level variations. The correlations between D and clinical parameters for COPD were examined and compared to a conventional emphysema index (EI%). 100 COPD patients underwent CT scans at full inspiration and expiration. D showed strong agreement between breath-hold levels, while EI% did not. D exhibited significant correlations with airflow limitation, diffusion capacity, exercise capacity, and quality of life. The transfer of this research idea to Coreline Soft for commercialization highlights its potential as a valuable tool for assessing emphysema severity, minimizing the impact of breath-hold variations.

This study aimed to compare air trapping in healthy volunteers and asthmatics using pulmonary function tests and quantitative data from paired inspiratory and expiratory CT scans. ΔSV, representing the ratio of air fraction released after exhalation, was measured and used to calculate air trapping indexes. The study included 16 healthy volunteers and 9 asthmatics. The results showed significantly higher mean ΔSV, ΔSV0.4, and ΔSV0.5 values in asthmatics compared to healthy volunteers. Asthmatics also exhibited higher ΔSV values in the posterior lung. Air trapping indexes correlated negatively with clinical parameters in asthmatics. The ΔSV maps revealed abnormal geographic patterns in some asthmatic patients. The study concluded that quantitative assessment of ΔSV can differentiate the extent of air trapping between healthy volunteers and asthmatics. AVIEW COPD was used for data analysis in this study.

This study aimed to compare an integral-based half-band method (IBHB) with a conventional full-width half-maximum (FWHM) method for measuring peripheral airway dimensions in COPD. The IBHB was validated using airway phantoms and applied to 50 COPD patients. Airway parameters were measured and compared between the two methods. Significant differences were found in thin-section CT measurements, with IBHB yielding smaller wall area percentage and wall thickness. Correlations with functional parameters (FEV1, FEV1/FVC, GOLD stage) were stronger with IBHB. Linear regression analysis showed that fifth-generation WA% measured by IBHB was an independent variable, improving predictability when combined with the emphysema index. IBHB offers a different approach to measuring airway dimensions and shows better correlations with functional parameters in COPD. The method has been transferred to AVIEW COPD by Coreline Soft.

This study aimed to determine the optimal threshold for quantifying air trapping using the subtraction method on co-registered CT in COPD patients and correlate it with pulmonary function parameters. A total of 174 patients were included, and inspiration and expiration CT scans were performed, followed by non-rigid registration. The subtraction value per voxel between the scans was obtained, and the volume fraction of air trapping (air trapping index, ATI) was calculated using various thresholds. ATI, along with other parameters, was correlated with FEF25–75% and RV/TLC. The results showed that the highest correlation coefficients with FEF25–75% and RV/TLC were achieved at a threshold of 60 HU. These correlation coefficients were comparable to those obtained with other methods such as E/I MLD and Exp−856. The findings suggest that a threshold of 60 HU is optimal for quantifying air trapping and shows similar correlations with pulmonary function parameters. The idea of this research has been transferred to AVIEW COPD by Coreline Soft.

The purpose of this study was to develop and validate a semi-automatic segmentation method for thoracic cavity volumetry and mediastinal fat quantification in patients with chronic obstructive pulmonary disease (COPD). The method involved segmenting multiple organs, including the rib, lung, heart, and diaphragm, to separate the thoracic cavity region. A three-dimensional surface-fitting method was used to model the inner thoracic wall and diaphragm, taking into account lung disease-induced variations. The accuracy of the algorithm was evaluated by comparing its results to manual segmentations performed by two expert radiologists. The proposed method was also compared to three other segmentation methods using various evaluation metrics. The results showed that the proposed method achieved high accuracy, outperforming the other methods in terms of volumetric overlap ratio, surface distances, and false positive/negative ratios. The study concluded that the semi-automatic segmentation method can accurately extract multiple organs within the thoracic cavity and may have clinical utility. The method has been implemented in AVIEW Lung Lobe Segmentation by Coreline Soft.

This study introduces a fully automated method for accurately separating the left and right lungs using volumetric computed tomography (CT) scans. The method utilizes free-formed surface fitting and iterative 3-dimensional morphological operators to divide the lungs. A separating surface is determined by detecting a point set that traverses between the initial left and right lungs, and a free-formed surface-fitting algorithm models the separating surface. The left and right lung volumes are then separated using this surface. The performance of the automated method was evaluated against manual segmentation by a human expert using 44 CT examinations. The results demonstrated high accuracy, with small surface distances and volumetric overlap errors between the automatic and manual methods. The study validates the feasibility of automatically separating the left and right lungs through the identification of a continuous 3D separating surface in volumetric chest CT images. The proposed method has been implemented in the AVIEW Lobe Segmentation page.

This study aimed to evaluate air trapping index (ATI) measured by CT in COPD patients and nonsmokers, and its association with pulmonary function test (PFT) parameters. The study included 138 COPD patients and 29 nonsmokers. Results showed that COPD patients had higher ATI values in the whole lung and specific lung lesions compared to nonsmokers. ATI demonstrated stronger correlations with PFT parameters such as FEF25-75%, RV, and RV/TLC compared to the emphysema index (EI). ATI also showed similar correlations to other CT parameters like gas trapping index and expiration/inspiration ratio of mean lung density. Additionally, ATI in emphysema and hyperinflated areas on CT had better correlations with PFT parameters than the normal lung area. The study suggests that analyzing density changes in different lung areas using CT can provide valuable insights into the pulmonary functional impairment seen in COPD patients. The analysis method have been comercialized in AVIEW COPD, Coreline Soft

This review highlights advancements in radiology since the discovery of X-rays, including ultrasound, CT, MRI, digital imaging, and molecular imaging. It also discusses emerging trends such as imaging genomics and robotic interventions.

This study aimed to investigate the effect of using different CT scanners on the accuracy of high-resolution CT (HRCT) images in classifying regional disease patterns in patients with diffuse lung disease. The study used two scanners, GE and Siemens, and applied SVM and Bayesian classifiers to multicenter data. The results showed that the SVM classifier achieved better classification accuracy than the Bayesian classifier for each scanner. The use of an integrated dataset along with the SVM classifier had benefits in terms of classification accuracy of HRCT images acquired with more than one scanner. The method used in this article has been commercialized in AVIEW Lung Texture, Coreline Soft.

This paper introduces a novel approach to classify pulmonary arteries and veins from volumetric chest computed tomography (CT) images using AVIEW software. The authors construct a tree structure from vascular points while minimizing the construction cost using the vascular geometries and features of CT images. Their approach can handle non-contrast CT images as well as vascular contrast-enhanced images. The suggested approach shows clinically acceptable results in terms of classification capability, making it potentially useful in assessing many pulmonary diseases, such as pulmonary hypertension, interstitial lung disease, and COPD. The study used mathematical virtual phantoms and ten COPD noncontrast volumetric chest CT scans with submillimeter thickness for validation.

This study aimed to determine the impact of density correction on emphysema quantification and identify the optimal site for air density correction on volumetric CT scans. CT scans of 78 COPD patients were analyzed. Density correction using tracheal and external air measurements was conducted, and emphysema index (EI) and mean lung density (MLD) were calculated. Significant differences were observed in air densities, MLD, and EI before and after density correction. Correlations between CT quantification indices and pulmonary function tests (FEV1 and FEV1/FVC) improved after density correction. Tracheal air correction yielded better results than external air correction. The study concludes that density correction improves the correlation between emphysema quantification and pulmonary function tests on volumetric CT data.

The study aimed to assess whether contrast-enhanced dual-energy computed tomography angiography (DECTA) can be used for simultaneous evaluation of emphysema quantification and regional perfusion. The study evaluated 27 patients with pulmonary emphysema, and the results showed moderate correlations between quantitative parameters on virtual non-contrast (VNC) images and pulmonary function test data, and also moderate correlations between the severity of parenchymal destruction from VNC images and perfusion status from iodine maps. The study demonstrated that contrast-enhanced DECTA can be used for emphysema quantification and regional perfusion evaluation using VNC images and iodine maps, simultaneously. AVIEW software was used for lung segmentation and generation of the VNC color-coded map images.

A hierarchical support vector machine (SVM) with class-specific feature selection was proposed to enhance the accuracy and reduce the classification time for differentiating diffuse interstitial lung disease in computer-aided quantification. This method allowed each binary classifier to use a class-specific quasi-optimal feature set, and a computational cost-sensitive group-feature selection criterion was applied for accelerating the classification time. Compared to one-against-all and one-against-one SVM methods with sequential forward selection, the proposed method reduced classification time by up to 57% and significantly improved overall accuracy (paired t-test, p<0.001). These results suggest potential for the proposed method in real-time and online image-based clinical applications. This study also had an impact on the development of AVIEW Lung Texture software.

The study proposes the use of a context-sensitive support vector machine (csSVM) to enhance the identification of diffuse interstitial lung disease (DILD) in high-resolution computerized tomography (HRCT) images. The csSVM simultaneously uses the decision value of each class and information from neighboring regions to classify an ROI, resulting in significantly higher accuracy for ROI and whole lung classification than the conventional SVM classifier. The proposed method characterizes ROIs using 21 textual and shape features and was validated using 900 ROIs with fivefold cross-validation and 84 whole lung images categorized by a radiologist. The study concludes that the csSVM provides better overall quantification of DILD. Additionally, the study mentions that it affected the development of AVIEW Lung Texture, which may imply that the csSVM was implemented or used as a basis for the software's development.

This study aimed to investigate the correlation between CT-derived airway wall measurements and clinical indices in stable asthma patients taking inhaled corticosteroids. The study found significant correlations between several CT parameters, including wall thickness percentage (WT%), wall thickness (WT), bronchial-to-arterial diameter (BA) ratio, airway collapsibility (AC), and air trapping index (ATI), with post-bronchodilator FEV1/FVC, a noninvasive marker of airway remodeling. Additionally, WT, WT%, and WA% were found to be negatively correlated with lung function tests. The study suggests that WT%, WT, BA ratio, and AC on inspiration and expiration CT can be reliable indices for measuring airway remodeling in stable asthma patients. The airway measurement method employed in this study was transferred to AVIEW COPD software, which offers a convenient and effective way to analyze airway remodeling in patients with stable asthma.

This study aims to improve the performance of differentiating obstructive lung diseases based on high-resolution computerized tomography (HRCT) images by introducing new shape features and optimizing the classifier. The study utilized 265 HRCT images from 82 subjects, and two experienced radiologists selected regions of interest (ROIs) representing areas of severe centrilobular emphysema, mild centrilobular emphysema, bronchiolitis obliterans, or normal lung. In addition to 13 textural features, 11 shape features were employed to evaluate their contribution. The Bayesian classifier and support vector machine (SVM) were implemented, and a five-folding method was used to assess cross-validation of the system. The study found that adding shape features to conventional texture features significantly improved the overall sensitivity of both the Bayesian and SVM classifiers, particularly in smaller ROIs. The results suggest that shape features contribute more to overall sensitivity in smaller ROI sizes. The findings of this study have implications for the development of AVIEW Lung Texture, as it highlights the usefulness of adding shape features to conventional texture features for improving classification performance of obstructive lung diseases.

This paper presents a study on the performance of several machine classifiers for differentiating obstructive lung diseases using texture analysis on various region of interest (ROI) sizes in 265 high-resolution computerized tomography (HRCT) images taken from 92 subjects. Four machine classifiers were implemented: naïve Bayesian classifier, Bayesian classifier, artificial neural net (ANN) and support vector machine (SVM). The SVM showed the best performance in overall accuracy for classifying obstructive lung diseases in ROI sizes of 32x32 and 64x64, while the naïve Bayesian method performed significantly worse than the other classifiers. The study highlights the importance of selecting an appropriate classification scheme for improving performance based on the characteristics of the data set. This research has implications for the development of AVIEW Lung Texture, as it demonstrates the effectiveness of machine classifiers for automating quantitative analysis and avoiding intra-inter-reader variability in the diagnosis of obstructive lung diseases using texture analysis.

This study aimed to evaluate the correlation between CT parameters and clinical parameters in 34 patients with chronic obstructive pulmonary disease (COPD). The CT parameters included the extent of emphysema, air trapping, and airway thickening, which were measured using in-house software. The results showed that the CT parameters were correlated with pulmonary function test (PFT) results, body mass index (BMI), the modified Medical Research Council Dyspnea scale (MMRC scale), the six-minute-walk distance (6MWD), and the BODE index. Based on the findings, the commercial product, AVIEW COPD is developed by applying the analysis method used in this study. This software may provide a useful tool for evaluating the severity of COPD components and could potentially be used for clinical decision-making.

The study aimed to evaluate the impact of various factors, including the size of the airway and CT imaging parameters, on the automatic measurement of airway dimensions. An airway phantom containing tubes of varying diameters and wall thicknesses was used, and CT images were reconstructed with different kernels, field-of-view, and slice thicknesses. The results showed that accuracy was most reliable using images reconstructed with a standard kernel. The study concluded that standardization of imaging protocols and selecting the appropriate anatomical level are crucial for accurate measurement. The developed method was applied to AVIEW COPD's airway measurement algorithm.