DESIGN AND IMPLEMENTATION OF A DEEP LEARNING-BASED CAD SYSTEM FOR BREAST CANCER DETECTION IN ULTRASOUND SCAN

Breast cancer remains one of the most prevalent and life-threatening diseases affecting women globally, underscoring the urgent need for early and accurate diagnostic tools. While mammography is widely used for screening, its sensitivity is limited in patients with dense breast tissue, often necessitating supplementary imaging. Ultrasound (US) scans offer a non-invasive, cost-effective, and radiation-free alternative for detecting breast abnormalities. However, manual interpretation of ultrasound images is time-consuming and prone to variability, highlighting the demand for automated solutions. This study addresses this gap by designing and implementing a deep learning-based computer-aided diagnosis (CAD) system to enhance the accuracy and efficiency of breast cancer detection in ultrasound scans. The proposed system integrates advanced image preprocessing techniques with a convolutional neural network (CNN) architecture to automate the analysis of ultrasound scans. The design phase focuses on optimizing the CNN model to segment regions of interest and classify lesions as benign or malignant. During implementation, a curated dataset of breast ultrasound scans is preprocessed to address noise and artifacts, followed by augmentation to improve model generalizability. The CNN is trained end-to-end, leveraging transfer learning to enhance performance with limited data. The system’s effectiveness is rigorously evaluated using metrics such as accuracy, sensitivity, specificity, and area under the ROC curve (AUC), benchmarked against radiologist annotations and existing methods. Experimental results demonstrate that the developed CAD system achieves 95.2% accuracy and 93.8% sensitivity in detecting malignant lesions, outperforming traditional machine learning approaches like SVM. The implementation highlights the model’s ability to reduce false positives while maintaining robustness across diverse ultrasound scan qualities. This work not only validates the feasibility of deep learning-driven CAD systems in breast cancer diagnosis but also provides a scalable framework for deployment in clinical settings. Limitations, including dataset size and computational demands, are critically discussed, with proposed future directions such as integrating multimodal imaging (e.g. MRI or mammography) and adopting lightweight architectures for real-time processing. By bridging technical innovation with clinical utility, this project contributes to the growing body of research on AI-driven healthcare tools, offering a potential pathway to improve early diagnosis and patient outcomes in resource-constrained environments.