Nihon Hoshasen Gijutsu Gakkai zasshi · 2026
[Clinical Evaluation of AI-based Automatic Contouring for Thoracic Structures].
Аннотация на английскомОценочное исследованиеСтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Miura Hideharu, Nakao Minoru, Ochi Masanori, Kenjo Masahiro, Ozawa Shuichi, Kagemoto Masayuki
Аннотация
PURPOSE: This study aimed to verify the contouring accuracy of the artificial intelligence (AI)-based auto-segmentation software Contour+ (MVision AI Oy, Helsinki, Finland) both quantitatively and visually, and to evaluate its clinical validity for the thoracic region in Japanese patients.
METHODS: Ten thoracic radiotherapy cases with lung lesions were analyzed. Contour+ was used to automatically delineate both lungs, trachea, bronchus, esophagus, spinal cord, and heart. Three observers visually evaluated the auto-contours using a five-point scoring system, and the final manually corrected contours were used as the reference to calculate the dice similarity coefficient (DSC), Hausdorff distance (HD), and volume differences.
RESULTS: In all cases, the AI auto-contours were evaluated as "clinically acceptable with minor modifications (score ≥3)," with an average score of 4.4. The mean DSC values were 1.00 for the lungs, 0.99 for the trachea, 0.91 for the bronchi, 0.86 for the esophagus, 0.99 for the spinal cord, and 0.99 for the heart, indicating high agreement. The mean HD values were 5.68 mm, 8.72 mm, and 3.30 mm for the bronchi, esophagus, and heart, respectively. The mean volume changes after manual correction were 4.02 cc for the bronchi, 2.55 cc for the esophagus, and 2.58 cc for the heart.
CONCLUSION: AI-based auto-segmentation software Contour+ demonstrated high geometric agreement and clinical validity for major thoracic organs in Japanese patients, suggesting its potential to reduce the contouring workload and promote standardization in radiotherapy treatment planning.
METHODS: Ten thoracic radiotherapy cases with lung lesions were analyzed. Contour+ was used to automatically delineate both lungs, trachea, bronchus, esophagus, spinal cord, and heart. Three observers visually evaluated the auto-contours using a five-point scoring system, and the final manually corrected contours were used as the reference to calculate the dice similarity coefficient (DSC), Hausdorff distance (HD), and volume differences.
RESULTS: In all cases, the AI auto-contours were evaluated as "clinically acceptable with minor modifications (score ≥3)," with an average score of 4.4. The mean DSC values were 1.00 for the lungs, 0.99 for the trachea, 0.91 for the bronchi, 0.86 for the esophagus, 0.99 for the spinal cord, and 0.99 for the heart, indicating high agreement. The mean HD values were 5.68 mm, 8.72 mm, and 3.30 mm for the bronchi, esophagus, and heart, respectively. The mean volume changes after manual correction were 4.02 cc for the bronchi, 2.55 cc for the esophagus, and 2.58 cc for the heart.
CONCLUSION: AI-based auto-segmentation software Contour+ demonstrated high geometric agreement and clinical validity for major thoracic organs in Japanese patients, suggesting its potential to reduce the contouring workload and promote standardization in radiotherapy treatment planning.