Integrating Transfer Learning and Flower Pollination Algorithm for Breast Cancer Histopathology Image Classification

Breast cancer classification through histopathology images plays an essential role in early diagnosis and treatment planning. Manual examination of tissue slides can lead to inconsistencies due to complexity and visual similarity between classes. This project introduces a hybrid approach combining deep learning and nature-inspired optimization to improve classification accuracy. Three pre-trained convolutional neural network models—MobileNet, DenseNet, and EfficientNet—are employed for feature extraction using transfer learning. To refine these features, the Flower Pollination Algorithm (FPA) is used as an optimization strategy to reduce redundancy and highlight informative patterns. For the classification task, Random Forest and Support Vector Machine (SVM) models are applied based on the extracted and optimized features. A web interface built with Flask allows users to upload images, process data, and view classification results. This system aims to enhance diagnostic precision by integrating feature-rich representations and effective feature selection. Performance is evaluated based on accuracy, precision, recall, and F1-score. The combination of deep models and FPA contributes to a robust and efficient classification framework suitable for histopathology image analysis.

Keywords: Breast cancer, Histopathology, Transfer learning, MobileNet, DenseNet, EfficientNet, Flower Pollination Algorithm, Random Forest, SVM, Flask.