Classification of 3D Dynamic Contrast Enhanced Breast Tumour MRI Images Using CNN

Breast cancer remains the most prevalent malignant tumor in women worldwide and a significant contributor to female mortality. Dynamic contrast-enhanced MRI (DCE-MRI) is an emerging technique for the diagnosis of breast disease, and it is highly efficient in the recognition, observation, and diagnosis of tumors. Additionally, there is hope that more advanced applications for early breast cancer diagnosis may be developed by recent advances in image processing and deep learning. In our study, a 3D CNN deep learning tool was designed to classify breast tumors into stages of malignancy in contrast to prior works that generally classify images as benign or malignant. Firstly, the 3D DCE-MRI images are preprocessed and then split into training and validation sets. The train set was augmented, and combined data was fed into 3D CNN, which provided 81% training accuracy and 71% validation accuracy. Further work, a prediction application program interface was created using our pretrained model in which the stages for each of the 3D images loaded were predicted. The hypothesized 3D CNN model has shown good accuracy in classifying the four distinct stages of breast tumors, namely stage 1, stage 2, stage 3, and stage 4. This can aid radiologists in making more accurate predictions and in the early diagnosis of breast tumors.