Evaluating Self-Supervised Learning Models for Multi-Label Pathology Detection: A Benchmark Against Supervised Learning

Evaluating Self-Supervised Learning Models for Multi-Label Pathology Detection: A Benchmark Against Supervised Learning

연구 분야: Artificial Intelligence

학회: 2025 2nd International Conference on Computational Intelligence, Communication Technology and Networking (CICTN)

The analysis and objective of the current work is to check the performance of SSL Models in multi-label classification on the available dataset of NIH Chest X-ray images which is a large-scale medical imaging dataset. The performance of SSL models, Dino_v_bl6, Dino with ResNet-50, and MoCo (Momentum Contrast) has been evaluated against the supervised model ResNet. These models are supposed to train to detect multiple pathologies using chest X-ray images. We also note that self-supervised embedding models outperform traditional supervised learning to a significant extent, achieving similar or higher performance on a number of quantitative metrics when compared to the softmax supervised ResNet model. This also draws attention to the possibilities of using SSL for biosignal-specific medical images. Furthermore, SSL approaches offer a promising solution in scenarios where there is an overabundance of annotated datasets available for training models such as deep neural networks. The effectiveness of SSL approaches in addressing the problem of data annotation is also emphasized by this study.