การประเมินความแม่นยำของเครื่องมืออ่านภาพรังสีทรวงอกอัตโนมัติปัญญาประดิษฐ์(Artificial Intelligence; AI) ในการแปลผลวัณโรคปอดจากภาพรังสีทรวงอกในผู้ป่วยนอกที่มีอาการสงสัยวัณโรคปอด อ.แม่ระมาด จ.ตาก
Year: 2021
Type: article
Abstract: Background: Chest radiograph (CXR) has been an important method for tuberculosis screening. However, there were limitations in CXR interpretation. Using Artificial Intelligence technology (AI) to read CXR for TB detection has been introduced. Objectives: To evaluate the diagnostic accuracy of AI detection of pulmonary TB using AFB and clinical diagnosis as the reference standard.Methods: The study was a diagnostic test study of AI from INET company, Thailand. The data was collected retrospectively in an electronic medical record for those who had CXR and AFB tests together with having any TB suggestive symptoms from February - September 2020. The study defined Sensitivity, Specificity, Positive Predictive Value(PPV), Negative Predictive Value (NPV), and Area Under the Curve (AUC) at a cutoff score of 70 out of 100 and defined the cutoff score that would make the sensitivity over 90%Results: For AFB as the reference standard, AI has sensitivity 89.65%, specificity 74.38%, PPV 16.25%, NPV 99.23% and AUC 0.888 (95% CI 0.827-0.949). For clinical diagnosis as the reference standard, AI has Sensitivity 85.42%, Specificity 76.39%, PPV 25.62%, NPV 98.21% and AUC 0.885 (95% CI 0.842-0.928). For Sensitivity over 90%, the cutoff score should be at 38-54 out of 100.Conclusion: AI offered good diagnostic accuracy in CXR interpretation for tuberculosis detection in patients with TB suggestive symptoms.
Author: มหัสฉริยพงษ์, พัชริดา
Language: Thai
Subfield: Infectious Diseases
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7219962212
