Transfer Learning for Classification of Histopathology Images of Invasive Ductal Carcinoma in Breast

Breast cancer is a leading cause of death in women all over the world. It is critical to diagnose it early in order to begin treatment. For the past decade, researchers have used machine learning algorithms to analyze medical images. The Deep learning concepts helped a lot in the medical field to detect the pattern, classification, etc. It aids in early cancer detection, helping us to save lives. In this paper, we talked about Invasive Ductal Carcinoma(IDC). With the improvement in the field of the neural network, this research work explores the use of the convolutional neural networks to detect IDC(+ve or -ve). The dataset used is from Kaggle and contains 277,524 total images of size 50x50. 198,738 images are non-cancerous and 78,786 images are cancerous. We proposed a transfer learning method for our classification of IDC negative(non-cancerous) and IDC positive (cancerous). Versions of Resnet, Efficientnet, and Densenet were used in this research. Resnet101 model is the one with the highest accuracy of 93.14% and an AUC of 0.9494 outperformed the other models we used.