What question did this study set out to answer?

The aim is to develop an AI-based system that improves breast cancer diagnosis through deep learning and interpretability techniques.

April 3, 2026

Onco Breast Cancer Detection System: Explainable Deep Learning Framework for Ultrasound-Based Breast Cancer Diagnosis

Key Points

The aim is to develop an AI-based system that improves breast cancer diagnosis through deep learning and interpretability techniques.
Utilized a convolutional neural network to analyze ultrasound images.
Trained using the BUSI Breast Ultrasound dataset for classification of benign, malignant, and normal images.
Integrated Grad-CAM for visualizing impactful regions within ultrasound images.
Provided a real-time dashboard for clinicians to upload images and generate diagnostic reports.
Achieved high classification accuracy across evaluation metrics such as precision, recall, and F1-score.
Grad-CAM explanations enhanced trust in AI-assisted diagnoses by providing interpretable insights.
Demonstrated effective support for clinicians in early breast cancer detection.

Abstract

Breast cancer remains one of the most common and life-threatening diseases affecting women worldwide. Early detection significantly increases survival rates and improves treatment outcomes. However, manual diagnosis using ultrasound images requires experienced radiologists and may lead to diagnostic variability. This paper proposes an AI-driven breast cancer detection system that combines deep learning classification with explainable artificial intelligence techniques to assist medical professionals in early diagnosis. The proposed system utilizes a Convolutional Neural Network (CNN) trained on the BUSI Breast Ultrasound dataset to classify images into three categories: benign, malignant, and normal. To improve model transparency, Grad-CAM explainability is integrated to visualize the regions of ultrasound images that influence the model's prediction. A real-time Streamlit dashboard enables clinicians to upload ultrasound images, obtain predictions, visualize heatmaps, and generate downloadable diagnostic reports. Experimental results show that the proposed model achieves high classification accuracy and strong performance across evaluation metrics including precision, recall, and F1-score. Grad-CAM explanations provide interpretable insights that improve trust in AI-assisted diagnosis. The system demonstrates how combining deep learning with explainable AI can support clinicians in early breast cancer detection and improve healthcare decision- making.

Mark Helpful

Bookmark

Relay

View Full Paper