35. Ứng dụng bước đầu trí tuệ nhân tạo Qure.AI X-quang lồng ngực trong chẩn đoán lao phổi

Abstract

Tuberculosis is still a global health concern with yearly increased number of people diagnosed with pulmonary TB. Therefore, the issue of community-based TB screening and early diagnosis of pulmonary TB is an urgent need. Artificial intelligence was applied for the diagnosis of pulmonary tuberculosis to target early detection of lesions and accurate diagnosis. The purpose of our study was to initially apply Qure.AI for chest X-ray in the diagnosis of pulmonary tuberculosis. The study described 126 patients suspected for pulmonary TB; Qure.AI read their chest Xrays and microbiological or histopathological tests were conducted to diagnose pulmonary TB. The final diagnosis were compared with AI. We found a good consensus between the 2 methods. AI sensitivity was 88.8%, specificity was 48.8%, predictive value was negative 71%. The area under the ROC curve was 77.1% with p < 0.001, 95%CI: 0.69 - 0.86. Thus, the chest X-ray images read by Qure.AI were valuable in diagnosing pulmonary tuberculosis with a fairly good level of accuracy with cut-off score of 0.503. Our research showed that AI had the potential to become an essential tool in supporting early diagnosis of pulmonary TB, and community lung TB screening.