Performance analysis of CNN models for evaluating Biglycan as a biomarker in predicting breast cancer

Breast cancer is found in the present generation at a higher level. Since the middle of the 20th century numerous unfavorable changes have been made to food, our way of life, obesity, the situation, and our microbiomes. Because of these variations, people have become more susceptible to cancer at an earlier age. The considered dataset is images of breast cancer-positive and cancer-negative histological sections utilizing the Biglycan biomarker. The dataset and models created from it can be used by pathologists and other healthcare practitioners to address probable disorders. Three widely used CNN models were examined for performance, and we discovered that MobileNet-V2 outperformed them all. Transfer learning-based breast cancer classification with the Biglycan biomarker dataset helps in tracking at the earlier stages. Additionally, our model outperformed earlier research using the same dataset in terms of accuracy. Our ability to create high-performance devices with lower power consumption and cheaper costs is made possible by the MobileNet-V2's ability to achieve greater precision while using a lighter model. In comparison with the existing work using LeNet, MobileNet-V2 outperforms.