Cancer shield: Ai-powered breast cancer detection

Mehat, Siti Azalia (2024) Cancer shield: Ai-powered breast cancer detection. Project Report. Melaka, Malaysia, Universiti Teknikal Malaysia Melaka. (Submitted)

Text (Full Text) Cancer shield Ai-powered breast cancer detection.pdf - Submitted Version Download (3MB)

Abstract

This study explores the application of deep learning models for breast cancer detection using mammogram and ultrasound images, aiming to improve traditional tumor detection and classification methods. The project evaluates four models: VGG16, VGG16 (fine-tuned), ResNet50, (fine-tuned) and ResNet50, to identify the most accurate approach for tumor detection. We utilized the CBIS-DDSM dataset, which comprises 2,260 mammogram images divided into training, testing, and validation sets. Key evaluation metrics, including accuracy, processing time, loss, and precision, were employed to assess model performance. Additionally, the Breast Ultrasound Images Dataset, containing 1,578 images categorized into benign, malignant, and normal classes, was integrated into the study. The inclusion of the normal class was critical, as it provided a comparative baseline for identifying non-cancerous cases, enhancing the robustness of the models. Our results demonstrate significant advancements in diagnostic medicine, reducing the need for extensive manual diagnoses. The fine-tuned ResNet50 model achieved the highest accuracy at 97%, outperforming the VGG16 (fine-tuned) model at 92%, ResNet50 at 88%, and VGG16 at 84%. These findings underscore the potential of advanced deep learning techniques in early breast cancer detection, offering a more efficient solution for healthcare professionals.

Actions (login required)

View Item