Enhancing breast cancer detection accuracy through machine learning, deep learning and transfer learning techniques for clinical practice

Abstract Breast cancer is a pervasive global health concern, impacting millions of women worldwide. Timely detection and precise diagnosis are pivotal factors in improving patient outcomes. This review presents a comprehensive analysis of machine learning (ML), deep learning (DL), and transfer learning (TL) models applied to breast cancer detection. It encompasses the classification of different types of breast cancer, prognosis, diagnosis, prediction, and clinical decision support. The present study examines a wide range of articles to recognize the frequently used architectures, datasets, activation functions, and evaluation metrics. Furthermore, the review scrutinizes the effectiveness of various AI techniques in predicting and diagnosing breast cancer, elucidating various evaluation metrics and their utilization. The WDBC and BreakHis databases are image datasets commonly used for breast cancer prediction. The performance of machine learning, deep learning, and transfer learning algorithms varies significantly in terms of precision, recall, F1 score, and accuracy. CNN model is the most commonly used deep learning technique, with the study indicating that it is used by about 60% of researchers. In terms of network architecture, ResNet is used by about 57% of researchers. Conspicuously, Softmax occurs as the most frequently used activation function i.e., 89%, and accuracy and precision are the foremost metrics for performance evaluation i.e., 60%. According to the study, deep learning and transfer learning methods achieve the highest accuracy, reaching 99.54% in breast cancer detection which raises concerns about dataset bias, overfitting, and lack of external validation. In terms of machine learning based breast cancer detection, the random forest algorithm demonstrates remarkable success, achieving the highest accuracy rate of 99%. This review serves as a comprehensive exploration of the current state of AI applications in breast cancer, highlighting their potential to reshape the landscape of breast cancer healthcare.