Multiclass Breast Cancer prediction from Mammogram using CNN

The breast cancer is a greatest threat among women across the globe. In the modern era of medical image analysis, the accurate identification of breast cancer from mammogram images is a critical task. This article proposes a methodology for developing a multiclass breast cancer classifier using deep learning architecture. The proposed methodology uses a sequential model architecture comprising Conv2D, MaxPooling2D, Flatten, Dropout and Dense layers. All these layers serve a specific role in feature extraction and classification. The model extracts appropriate features from mammogram images and progressively learns and discriminates representations to enhance classification accuracy. The incorporation of regularization techniques such as Dropout layers overcomes overfitting, improving the acquisition of robust features. ROC curve analysis is used in the compilation and training phases to evaluate the model performance. This article provides a systematic approach to the development and evaluation of a deep learning model for breast cancer identification. This also offers insights into its architectural framework, training process and performance assessment techniques. The proposed method has achieved classification accuracy of 98%.