An Efficient Methodology for the Classification of Invasive Ductal Carcinoma Using Transfer Learning

Abstract

Currently, breast cancer (BC) is one of the leading causes of death in women. It is still a challenging task for the pathologist to diagnose cancer properly. In this research study, the communal kind of BC that’s invasive ductal carcinoma (IDC) has been classified. For the classification of IDC, lots of developing techniques have been used in the medical research field. But many problems are faced regarding time to detect cancer cells in patients, class imbalance, data overfitting, less accuracy rate, and vanishing gradient in the BC classification technique. So, it is vital to develop an accurate, and well-organized system for the classification of IDC. To overcome these problems, an efficient methodology has been proposed in which we have defined the classification model as the TransResCNN model i.e. (transfer learning applied to the pre-trained residual network (ResNet) CNN model). Most popular transfer learning and data augmentation methods are applied for dealing with a huge dataset. For the performance evaluation, the model evaluates through a confusion matrix for image-based classification of IDC. Several evaluation metrics have also been applied like accuracy, precision, F1- score, and recall. A comparison of various existing studies performed with our proposed study shows that the proposed study achieved the highest accuracy of 90.76% and an F1-score of 93.56%. The investigated study shows that our proposed methodology achieved an improved performance than previous research studies.