Frontiers in molecular biosciences · 2026
External validation of a commercial AI system for pulmonary embolism detection on chest CTPA: a multicenter 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.
Tudi Mireayi, Abudoubari Saimaitikari, Waresi Xierenayi, Ainiwaer Aikebaierjiang, Abudouwufu Nuermaimaijiang, Wumaier Palidanmu, Abula Adilijiang, Xia Yuwei, Aisika Ailiyaerjiang, Qiu Ya, Tuerxun Maimaitiaili, Tuersun Abudouresuli
Аннотация
BACKGROUND: Pulmonary embolism (PE) is a critical cardiovascular emergency requiring prompt, accurate diagnosis. CT pulmonary angiography (CTPA) is the diagnostic gold standard, yet rising case volumes and radiologist shortages challenge clinical workflows. Artificial intelligence (AI) offers potential to enhance diagnostic precision and efficiency. This multicenter study validates the performance of a commercially available AI system compared with radiologist interpretation alone and in combination.
METHODS: In this retrospective analysis, 600 consecutive patients suspected of PE underwent CTPA between January 2024 and May 2025 at three hospitals in Xinjiang. All scans employed 256-slice CT with standardized protocols (100 kV, 0.625 mm slice thickness, iohexol contrast). Images were processed using uAIDiscover PE software, generating Pulmonary Thrombus Burden Score (PTBS). Manual Pulmonary Artery Obstruction Index (PAOI) was independently scored via the Qanadli system by consensus of three senior radiologists, serving as the reference standard. Diagnostic accuracy and correlation between AI and manual scores were assessed (SPSS 24.0; P < 0.05).
RESULTS: Among 600 patients analyzed, 271 (45.2%) had pulmonary embolism. PE patients had significantly higher BMI and greater prevalence of hypertension and coronary artery disease (P < 0.05). ROC analysis demonstrated superior diagnostic performance for the combined manual + AI approach across all centers (AUC: 0.928-0.934) compared to AI alone (AUC: 0.807-0.810) or manual reading alone (AUC: 0.888-0.914). AI processing was remarkably fast at 0.19 ± 0.02 min versus 5.26 ± 0.94 min for radiologists alone, while combined approach required 2.61 ± 0.69 min. Strong correlation was observed between AI-derived PTBS and manually calculated PAOI (r = 0.863, P < 0.001). The combined approach significantly reduced diagnostic errors to 7 cases compared to 43 for AI alone and 29 for manual reading alone.
CONCLUSION: Integration of AI with manual interpretation improves pulmonary embolism detection accuracy and reduces reading time, supporting its implementation to optimize clinical workflow and patient outcomes.
METHODS: In this retrospective analysis, 600 consecutive patients suspected of PE underwent CTPA between January 2024 and May 2025 at three hospitals in Xinjiang. All scans employed 256-slice CT with standardized protocols (100 kV, 0.625 mm slice thickness, iohexol contrast). Images were processed using uAIDiscover PE software, generating Pulmonary Thrombus Burden Score (PTBS). Manual Pulmonary Artery Obstruction Index (PAOI) was independently scored via the Qanadli system by consensus of three senior radiologists, serving as the reference standard. Diagnostic accuracy and correlation between AI and manual scores were assessed (SPSS 24.0; P < 0.05).
RESULTS: Among 600 patients analyzed, 271 (45.2%) had pulmonary embolism. PE patients had significantly higher BMI and greater prevalence of hypertension and coronary artery disease (P < 0.05). ROC analysis demonstrated superior diagnostic performance for the combined manual + AI approach across all centers (AUC: 0.928-0.934) compared to AI alone (AUC: 0.807-0.810) or manual reading alone (AUC: 0.888-0.914). AI processing was remarkably fast at 0.19 ± 0.02 min versus 5.26 ± 0.94 min for radiologists alone, while combined approach required 2.61 ± 0.69 min. Strong correlation was observed between AI-derived PTBS and manually calculated PAOI (r = 0.863, P < 0.001). The combined approach significantly reduced diagnostic errors to 7 cases compared to 43 for AI alone and 29 for manual reading alone.
CONCLUSION: Integration of AI with manual interpretation improves pulmonary embolism detection accuracy and reduces reading time, supporting its implementation to optimize clinical workflow and patient outcomes.