A Multi-Center Validation Study in Thai Population

View the PDF file from the paper entitled “Discovering deep breast cancer in learning in mammosis: Study the verification of the multiple center in the Thai population, by ISARUN CHAMVEHA and 14 other author

PDF HTML (experimental) view

a summary:This study provides a deep educational system for detection of breast cancer in the mammal imaging, which was developed using a modified NetV2 structure with improved attention mechanisms. The model was trained in X -ray breast imaging from the main Thai Medical Center and its authenticity has been validated on three distinct data settings: a group test in the field (9,421 cases), a confirmed group of biopsy (883 cases), and a general circular outside the scope (761 cases) collected from two different hospitals. To detect cancer, the Aurocs Form of 0.89, 0.96 and 0.94 achieved the relevant data sets. The ability of the regime’s lesion localization, which was evaluated using measures including the lesion localization part (LLF) and part of non -papillomatic localization (NLF), showed a strong performance in identifying suspicious areas. Clinical verification through compatibility tests showed a strong agreement with radiologists: 83.5 % rating, 84.0 % corresponding to resettlement for the biopsy, 78.1 % rating and 79.6 % settlement of resettlement for cases outside the field. Expert radiologists’ acceptance rate also reached 96.7 % in the confirmed cases of biopsy, and 89.3 % for cases outside the field. The system has achieved the scale of the system use of 74.17 for the source hospital, and 69.20 for health verification hospitals, indicating good clinical acceptance. These results show the effectiveness of the model in helping the interpretation of X -ray breast imaging, with the possibility of enhancing the functioning of breast cancer examination in clinical practice.