BMC musculoskeletal disorders · 2026
Evaluation of AI developed for rib fracture detection in identifying vertebral fractures on abdominal CT.
СтатьяОценочное исследованиеНазвание и аннотация: Перевод пока не подготовлен
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.
Kim Hyunjung, Gao Yongfeng, Kim Jin Woo, Ahn Jhii-Hyun, Yoon Ga-Young, Kim Yeon-Jun, Ko Sung Min
Аннотация
PURPOSE: To evaluate the performance of an AI algorithm originally developed for rib fracture detection in identifying vertebral fractures on abdominal CT scans.
METHODS: A retrospective study was performed on 58 patients with both CT and MRI scans. An AI model and ER clinicians independently assessed CT images for fractures of the vertebral body, transverse processes, and spinous processes. Two spine radiologists reviewed both CT and MRI data and established the reference standard.
RESULTS: At the patient level, both the AI algorithm and ER clinicians achieved 100% sensitivity, specificity, accuracy, and F1-score. At the vertebral level, the AI algorithm achieved higher sensitivity than ER clinicians (67% vs. 54%, p < 0.05), with similar accuracy (93% vs. 94%, p = 0.73). At the anatomical location level, the AI algorithm achieved a sensitivity of 62%, accuracy of 98%, and an F1-score of 0.69, compared with 46% sensitivity (p < 0.05), 97% accuracy (p = 0.92), and an F1-score of 0.62 (p = 0.50) for ER clinicians. Among 178 confirmed fracture locations, the AI algorithm detected 110 cases, whereas ER clinicians identified 81. Of the 97 fractures missed by ER clinicians, the AI algorithm correctly detected 51.
CONCLUSIONS: The AI software originally trained for rib fracture detection identified vertebral fractures with higher sensitivity compared to those detected by ER clinicians. Its ability to identify otherwise missed fractures highlights its value as a supplementary tool in urgent care.
METHODS: A retrospective study was performed on 58 patients with both CT and MRI scans. An AI model and ER clinicians independently assessed CT images for fractures of the vertebral body, transverse processes, and spinous processes. Two spine radiologists reviewed both CT and MRI data and established the reference standard.
RESULTS: At the patient level, both the AI algorithm and ER clinicians achieved 100% sensitivity, specificity, accuracy, and F1-score. At the vertebral level, the AI algorithm achieved higher sensitivity than ER clinicians (67% vs. 54%, p < 0.05), with similar accuracy (93% vs. 94%, p = 0.73). At the anatomical location level, the AI algorithm achieved a sensitivity of 62%, accuracy of 98%, and an F1-score of 0.69, compared with 46% sensitivity (p < 0.05), 97% accuracy (p = 0.92), and an F1-score of 0.62 (p = 0.50) for ER clinicians. Among 178 confirmed fracture locations, the AI algorithm detected 110 cases, whereas ER clinicians identified 81. Of the 97 fractures missed by ER clinicians, the AI algorithm correctly detected 51.
CONCLUSIONS: The AI software originally trained for rib fracture detection identified vertebral fractures with higher sensitivity compared to those detected by ER clinicians. Its ability to identify otherwise missed fractures highlights its value as a supplementary tool in urgent care.