DEVELOPMENT OF AN AI-POWERED EARLY BREAST CANCER DETECTION SYSTEM USING THERMAL IMAGING, EFFICIENTNETB3, AND SUPPORT VECTOR MACHINE CLASSIFICATION

NDUBUISI, OGECHI JOSEPHINE (2026) DEVELOPMENT OF AN AI-POWERED EARLY BREAST CANCER DETECTION SYSTEM USING THERMAL IMAGING, EFFICIENTNETB3, AND SUPPORT VECTOR MACHINE CLASSIFICATION. Other thesis, Godfrey Okoye University, Enugu Nigeria.

Text Ndubuisi_Ogechi_Josephine_Final-Project_MERGED.docx.pdf Download (7MB)

Abstract

Breast cancer remains one of the leading causes of cancer-related deaths among women worldwide, particularly in developing countries where access to early screening facilities is limited. This study was motivated by the need for a non-invasive, radiation-free, and accessible method for supporting early breast cancer detection using thermal imaging and artificial intelligence. To address this challenge, a hybrid EfficientNetB3–SVM model was developed for the classification of breast thermographic images as normal/benign or abnormal/malignant. Thermal images from the DMR-IR and Mendeley Breast Thermography datasets were combined, producing a total of 619 images that were preprocessed using contrast enhancement, resizing, normalization, and data augmentation techniques. EfficientNetB3 was employed for deep feature extraction, while a Support Vector Machine with an RBF kernel performed the final classification. The developed model was integrated into a full-stack web application named BreastCare AI, featuring user authentication, real-time prediction, Grad-CAM visual explanations, automated PDF report generation, scan history, and comparison functionalities. Experimental evaluation on the test dataset achieved an accuracy of 81.45%, sensitivity of 85.37%, specificity of 79.52%, and an AUC-ROC score of 0.87, while cross-validation demonstrated the model's robustness and reliability. The study contributes an explainable and deployable AI-powered breast cancer screening system that combines intelligent classification, visual interpretability, and patient monitoring within a single platform, thereby providing a practical decision-support tool for early breast cancer detection.

Item Type:	Thesis (Other)
Subjects:	Q Science > Q Science (General)
Divisions:	Faculty of Computing And Information Technology (FACIT)
Depositing User:	Cynthia Ugwuoti
Date Deposited:	24 Jul 2026 13:40
Last Modified:	24 Jul 2026 13:40
URI:	http://eprints.gouni.edu.ng/id/eprint/5989

Actions (login required)

View Item