The School of Computer Science is pleased to present…
Advanced Frameworks in Histopathology: Enhancing Breast Cancer Classification and Retrieval through Interactive Computational Techniques
PhD Dissertation Defense by: Abdala Nour
Date: Wednesday, August 13th, 2025
Time: 11:00 am
Location: Essex Hall, Room 122
The proposed research introduces the Interactive Computational Retrieval of Histopathology for Breast Cancer (ICRH-BC) framework, a sophisticated computational approach designed to enhance diagnostic precision in clinical pathology. Addressing challenges such as limited histopathological data and diagnostic ambiguity, the framework refines classification and retrieval processes for breast cancer tissue analysis. It integrates a Conditional Variational Autoencoder (CVAE) with self-attention mechanisms to enhance feature extraction and conditional image generation. In tandem, it incorporates a kernel-PCA-optimized content-based image retrieval (CBIR) pipeline, leveraging cosine similarity, Gabor filters, and dimensionality reduction to extract distinctive features and improve classification accuracy. These enhancements support the identification of breast cancer subtypes based on complex tissue morphology. Evaluated on an extensive dataset of histopathological images, the framework demonstrates high retrieval performance and diagnostic accuracy. The attention mechanisms contribute to improved context awareness by prioritizing relevant image features. The initial baseline CNN accuracy was 0.84 with an AUC of 0.73, which increased to 0.91 and 0.87, respectively, after integrating the attention-augmented ACVAE model, resulting in an improvement of 8.33% in accuracy and 19.18% in AUC. These results underscore the impact of advanced data augmentation and attention mechanisms in boosting model performance, positioning ICRH-BC as a promising tool for histopathological breast cancer classification.
Interactive learning, Kernel PCA, Cosine similarity, Gabor filters, Conditional Variational Autoencoder(cVAE), Attention mechanism, and Content-Based Image Retrieval (CBIR)
Internal Reader: Dr. Imran Ahmad
Internal Reader: Dr. Dima Alhadidi
External Reader: Dr. Mohamed Belalia
External Examiner: Dr. Faisal Qureshi, Computer Science, Ontario Tech University
Advisor(s): Dr. Boubakeur Boufama
Chair: Dr. John Trant, Department of Chemistry and Biochemistry
