A Study on the Clinical Effectiveness of Deep Learning CAD Technology

• Published in: 2022 International Conference on Artificial Intelligence in Information and Communication (ICAIIC) pp. 194 - 196
• Main Authors: Ju-Hyuck, Han, Hyun-Woo, Oh, Woong-Sik, Kim
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.02.2022
• Subjects: CAD; Chest Radiography; Computational modeling; Data models; Deep learning Algorithm; Observer Performance Test; Observers; Optimized production technology; Solid modeling; Training; Training data
• DOI: 10.1109/ICAIIC54071.2022.9722681

Chest radiography is the most common method of examining chest disease. However, interpretation of chest X-rays is difficult, and the diagnosis may vary depending on the doctor's proficiency. In order to solve this problem, additional diagnosis using a computer is attracting attention in the medical imaging field. In addition, the recently developed artificial intelligence technology has been applied to the analysis of chest X-rays, and commercialization has entered the stage as a computer-aided diagnostic tool. However, the reading model based on artificial intelligence has different performance depending on the type of data. In addition, current medical data is a weak standardization stage and the data form varies from institution to institution. Therefore, the performance of the model may not be guaranteed if the data for training artificial intelligence and the data from the real institution are different. The purpose of this study is to verify the clinical effectiveness of a computer-aided diagnostic tool based on chest X-rays. To this end, data from a different source than the training data were applied to the reading model. In addition, for validation, we prepared a doctor's lung lesion labeling findings for clinical validation. In this study, OPT (Observer Performance Test) was conducted by clinical experience level to evaluate the reading model.