A Comparative Study of Different Deep Learning Models Over Diverse Mammography Datasets
Sixiao Hu, Mou Zhou, Shangzhu Jin, Junsheng Peng, Yingxu Wang · 2024
In the field of computer-aided medical diagnosis, en-suring model robustness across diverse populations is essential for maintaining consistent performance and reliability. This paper examines the classification performance of several state-of-the-art image classification models, including EfficientNet_BO, Swin- Transformer, VisionTransformer, DenseNet121, ResNeXt, and Rep V gg, across multiple breast cancer datasets. We thoroughly evaluate these models to gauge their robustness against varying data distributions and demographic differences. The results of this study provide valuable insights into developing more robust and generalizable mammography diagnostic systems, ultimately enhancing diagnostic accuracy and improving healthcare out-comes. Our findings highlight both the strengths and limitations of each model, offering a solid foundation for future research and the optimization of breast cancer diagnostic tools in varied clinical settings.