2021/04/06 by Uğur Demir, Ismail Irmakci, Demir, Ugur +17 · 1 citation
Computer Science · Medicine · #AI in cancer detection #COVID-19 diagnosis using AI #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Radiomics and Machine Learning in Medical Imaging #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2104.02869
openalex publication_date 2021/04/06 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Visual explanation methods have an important role in the prognosis of the\npatients where the annotated data is limited or unavailable. There have been\nseveral attempts to use gradient-based attribution methods to localize\npathology from medical scans without using segmentation labels. This research\ndirection has been impeded by the lack of robustness and reliability. These\nmethods are highly sensitive to the network parameters. In this study, we\nintroduce a robust visual explanation method to address this problem for\nmedical applications. We provide an innovative visual explanation algorithm for\ngeneral purpose and as an example application, we demonstrate its effectiveness\nfor quantifying lesions in the lungs caused by the Covid-19 with high accuracy\nand robustness without using dense segmentation labels. This approach overcomes\nthe drawbacks of commonly used Grad-CAM and its extended versions. The premise\nbehind our proposed strategy is that the information flow is minimized while\nensuring the classifier prediction stays similar. Our findings indicate that\nthe bottleneck condition provides a more stable severity estimation than the\nsimilar attribution methods.\n