Selection of the best artificial intelligence techniques for analysis of gastrointestinal endoscopic images.

BACKGROUND AND STUDY AIM: Comprehensive identification and prioritization of artificial intelligence methods developed for the analysis of gastrointestinal endoscopic images can help in selecting the most appropriate techniques. This study aimed to introduce the best artificial intelligence techniques developed for analyzing gastrointestinal endoscopic images using fuzzy AHP-TOPSIS methods. PATIENTS AND METHODS: To identify the artificial intelligence techniques developed, a systematic search was conducted across five reputable databases. Subsequently, the Delphi method was employed to establish appropriate criteria for selecting the best artificial intelligence techniques. To estimate the relative weights of these criteria, the fuzzy analytical hierarchy process (FAHP) was utilized. Finally, to prioritize the identified artificial intelligence techniques, the technique for order preference by similarity to the ideal solution (TOPSIS) method was applied. RESULTS: 70 artificial intelligence techniques were identified. Seven selection criteria were introduced: validity, accuracy, comprehensiveness, processing time, cost, simplicity, and executive capability. The top methods selected were the computer-aided detection (CAD) system (0.8203), the human color appearance model (CIECAM) (0.8122), the combined method (PD-CNN-PCC-EELM) (0.8109), the combined method (DNN-CAD) (0.7928), and the ResNet18 deep learning model (0.7921), respectively. CONCLUSIONS: These findings represent a comprehensive approach to the developed techniques, which can be utilized to design methods with improved performance in the future.