Screening of Lung Cancer using Chest Radiographs with Application AI Chest for All (DMS TU) in the Context of a Regional Cancer Hospital

Abstract

Background : Lung cancer is one of the most common types of cancer and there is an increase in deaths from lung cancer every year. Although there is a lung cancer screening policy as a solution, it still has some problems. The shortage of radiologists can affect the effectiveness of lung cancer screening. According to the problems, AI Chest for All (DMS TU) has been developed to reduce radiologists’ workload of interpreting chest radiographs.Objective : to assess the qualifications of the AI Chest for All (DMS TU) application in the chest radiographic lung cancer screening.Method : Using a retrospective descriptive study, the samples in this study were the randomized chest radiographs of Udon Thani Cancer Hospital’s patients. Between January 1, 2018, and December 31, 2019, a total of 1,250 photos were taken and the instrument qualification was analyzed to determine whether AI Chest for All (DMS TU) is suitable to be used as a screening tool for cancer or not. This was compared with the method of interpretation of chest radiographs for the original lung cancer screening which was interpreted by expert radiologists.Results : AI Chest for All (DMS TU) contain 76.4% (95% CI=73.5% to 79.0%) of sensitivity, 89.3% (95% CI=85.2% to 92.6%) of specificity, 79% of accuracy, and 83% (95% CI 81% to 85%) of the area under the ROC curve (AUC). Comparing with the chest radiographic interpretation of the original lung cancer screening, which was interpreted by specialist radiologists, AI Chest for All (DMS TU) was less sensitivity but more specificity, accuracy and AUC.Conclusion : According to the above results, it can be concluded that AI Chest for All (DMS TU) is appropriate to be used as an alternative tool for interpreting chest radiographs of lung cancer screening.