The Classification of Breast Cancer Using a Transfer Learning Strategy in a Federated Learning Framework

In order to make an accurate forecast, deep learning algorithms need a sizable quantity of data to learn from. Recent research has shown that transfer learning-based DL techniques to developing CAD systems function properly across a range of use cases. Diseases including lung cancer, brain tumours, and breast cancer are detected and analysed at an early stage by employing these systems and their many modalities. Pre-trained models are often used for DL-based activities in computer vision instead of creating models of neural networks from scratch. This article explains why transfer learning models may be used to automate tumour classification without the need for augmentation or preprocessing. On the BreakHis dataset, seven machine learning models are used for tumour classification; Xception achieved the highest accuracy (83.07 ^% ) among these seven models. DarkN et53 also excels in computing a new measure called Balanced Accuracy (BAC) (87.17%), which is necessary for achieving accuracy with an imbalanced dataset. This discovery will help scientists and doctors choose the best model for tumour classification when faced with an imbalanced data set. It will help doctors categorise the illness more accurately and quickly.