Clinical imaging · 2025
AI-assisted chest radiograph interpretation enhances diagnostic confidence and standardizes diagnostic accuracy across radiologists: A multi-reader study.
СтатьяНазвание 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.
Huang Hao-Yu, Huang Yu-Han, Lin Cheng-Hsun, Tao Wan-Ting, Liao Wei-Chen, Yu Shuning, Mo Huei-Chi, Feng Wenyen, Hsu Yu-Ting, Wang Jian-Chiao, Ko Kai-Hsiung
Аннотация
PURPOSE: To evaluate the impact of an artificial intelligence (AI)-assisted computer-aided detection (CAD) system on the diagnostic accuracy and confidence in chest radiograph interpretation among nonthoracic radiologists and radiology residents with varying levels of experience.
METHODS: In this retrospective multiple-reader, multiple-case (MRMC) study, 400 chest radiographs (100 each for pulmonary nodules, pleural effusion, pneumothorax, and controls) were independently interpreted by 12 readers (two nonthoracic radiologists, four senior residents, and six junior residents). Readings were conducted under CAD-assisted and unassisted conditions, with a 30-day washout period. Readers assigned confidence scores (0-100) to their diagnosis. Diagnostic performance was evaluated using the area under the curve (AUC), sensitivity, and specificity, while reader confidence was assessed by the proportion of high-confidence ratings among correctly interpreted cases.
RESULTS: The AI-assisted CAD system improved diagnostic performance across all abnormalities, with significant gains for pulmonary nodules (AUC: 0.781 → 0.854; P < 0.001) and pleural effusion (0.896 → 0.948; P < 0.001). The sensitivity increased by 7.2% for effusion, while the specificity for nodules improved markedly by 15.7%. Among all the readers, junior residents showed the greatest gains, especially for nodules, where the CAD closed their baseline AUC gap (originally -7.3%, P = 0.006) relative to nonthoracic radiologists. Reader confidence also increased significantly with the CAD, particularly for nodules (+15.2 %; P < 0.001).
CONCLUSION: The AI-assisted CAD system significantly enhanced diagnostic accuracy and reader confidence in chest radiograph interpretation, especially for junior radiology residents. This approach may bridge experience-related diagnostic gaps and support clinical decision-making, particularly in institutions lacking thoracic radiologists.
METHODS: In this retrospective multiple-reader, multiple-case (MRMC) study, 400 chest radiographs (100 each for pulmonary nodules, pleural effusion, pneumothorax, and controls) were independently interpreted by 12 readers (two nonthoracic radiologists, four senior residents, and six junior residents). Readings were conducted under CAD-assisted and unassisted conditions, with a 30-day washout period. Readers assigned confidence scores (0-100) to their diagnosis. Diagnostic performance was evaluated using the area under the curve (AUC), sensitivity, and specificity, while reader confidence was assessed by the proportion of high-confidence ratings among correctly interpreted cases.
RESULTS: The AI-assisted CAD system improved diagnostic performance across all abnormalities, with significant gains for pulmonary nodules (AUC: 0.781 → 0.854; P < 0.001) and pleural effusion (0.896 → 0.948; P < 0.001). The sensitivity increased by 7.2% for effusion, while the specificity for nodules improved markedly by 15.7%. Among all the readers, junior residents showed the greatest gains, especially for nodules, where the CAD closed their baseline AUC gap (originally -7.3%, P = 0.006) relative to nonthoracic radiologists. Reader confidence also increased significantly with the CAD, particularly for nodules (+15.2 %; P < 0.001).
CONCLUSION: The AI-assisted CAD system significantly enhanced diagnostic accuracy and reader confidence in chest radiograph interpretation, especially for junior radiology residents. This approach may bridge experience-related diagnostic gaps and support clinical decision-making, particularly in institutions lacking thoracic radiologists.