Using hybrid pre-trained models for breast cancer detection.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Breast cancer diagnosis relies heavily on manual histopathology image analysis, which is labor-intensive and prone to inter-observer variability.
• Accurate and timely diagnosis is critical for effective breast cancer treatment and patient outcomes.
• Automated analysis of whole slide images (WSIs) offers a potential solution to improve efficiency and accuracy.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning model for the automated classification of invasive ductal carcinoma (IDC) in breast histopathology images.
• To compare the performance of the proposed model against existing machine learning and deep learning approaches.

Main Methods:

• A hybrid deep learning model combining Convolutional Neural Networks (CNNs) and EfficientNetV2B3 was developed.
• The model was trained and validated on whole slide images (WSIs) for the identification of IDC and non-IDC tissues.
• Performance was evaluated using metrics including accuracy, precision, recall, F1-score, MCC, AUC-ROC, and AUPRC.

Main Results:

• The proposed CNN+EfficientNetV2B3 model achieved high performance: 96.3% accuracy, 93.4% precision, 86.4% recall, 89.7% F1-score, 87.6% MCC, 97.5% AUC-ROC, and 96.8% AUPRC.
• The model significantly outperformed other tested deep learning models, including MobileNet+DenseNet121 and MobileNetV2+EfficientNetV2B0.
• The results indicate the model's robustness and superiority in classifying breast cancer tissues.

Conclusions:

• The hybrid CNN+EfficientNetV2B3 model demonstrates superior performance for automated breast cancer detection in histopathology images.
• This AI-driven approach can serve as a valuable tool to assist pathologists, enhancing diagnostic accuracy and efficiency.
• The findings suggest a promising advancement in computational pathology for breast cancer diagnosis.