A Novel Hybrid Approach for Breast Cancer Detection using Self-Supervised Learning and Machine Learning Models

Breast cancer is still one of the most common and fatal diseases in the world, which requires fast and precise diagnostic methods. Traditional deep learning models such as CNNs rely heavily on large labeled datasets, which can be challenging to obtain in medical imaging. In this study, a hybrid approach integrating Self-Supervised Learning (SSL) and ML techniques is proposed for breast cancer detection. The SSL-based SimCLR model is first trained on unlabeled histopathology images to extract meaningful feature representations. These features are then utilized to train traditional ML classifiers, including Random Forest, XGBoost, SVM, Logistic Regression, KNN, and Gradient Boosting for binary classification (benign vs. malignant). The proposed approach is evaluated on the BreakHis dataset, and the results demonstrate that SSL-based feature extraction combined with ML models outperforms traditional CNN-based classifiers such as Inception, ResNet50, and MobileNet. Among the ML classifiers, Random Forest achieved the highest performance with an accuracy of 82.5% and an AUC score of 0.893. The findings highlight the effectiveness of Self-Supervised Learning in improving breast cancer classification, reducing dependency on large labeled datasets, and enabling better generalization. This study suggests that SSL-based feature extraction with ML classifiers provides a promising alternative to fully supervised CNNs for breast cancer detection.