Award-winning research advances breast cancer image classification

Observer

27-04-2025

MUSCAT: Among the standout projects at the 11th National Research Award organised by the Ministry of Higher Education, Research and Innovation (MoHERI), the study titled 'Feed-forward networks using logistic regression and support vector machine for whole-slide breast cancer histopathology image classification' by Dr ArunaDevi Karuppasamy, Assistant Professor at the Department of Computing Sciences, Gulf College, received top honours in the Information and Communication Technologies field (PhD category).
Dr ArunaDevi's research addresses the critical challenge of histopathology image classification, essential for breast cancer diagnosis. While traditional methods rely on hand-crafted features, her study explores feed-forward approaches, proposing Convolutional Logistic Regression (CLR) and Convolutional Support Vector Machine (CSVM-H) networks.
Dr ArunaDevi Karuppasamy
These methods leverage predefined or externally learned filters, overcoming limitations of back-propagation techniques such as vanishing gradients and heavy data requirements.
Experiments showed that CLR and CSVM-H achieved superior accuracy and faster training times on small datasets from Sultan Qaboos University Hospital (SQUH) and BreaKHis, outperforming traditional models like VggNet-16 and ResNet-50. Both models demonstrated high Area Under Curve (AUC) scores (0.83 and 0.84 on the SQUH dataset), highlighting their efficiency for small, labelled datasets with limited computational resources.