A fully integrated computer-aided diagnosis system for digital X-ray mammograms via deep learning detection,...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Accurate breast mass detection, segmentation, and classification are crucial for effective computer-aided diagnosis (CAD) systems.
• Existing CAD systems often lack integration, requiring separate frameworks for different diagnostic stages.

Purpose of the Study:

• To propose and evaluate a fully integrated deep learning-based CAD system for digital X-ray mammograms.
• The system aims to enhance diagnostic accuracy by combining breast mass detection, segmentation, and classification within a single framework.

Main Methods:

• Utilized You-Only-Look-Once (YOLO) for initial breast mass detection.
• Employed a novel Full Resolution Convolutional Network (FrCN) for precise mass segmentation.
• Applied a Deep Convolutional Neural Network (CNN) for classifying masses as benign or malignant.

Main Results:

• Achieved high detection accuracy (98.96%) and F1-score (99.24%) for breast masses.
• Segmentation performance demonstrated high accuracy (92.97%) and Dice coefficient (92.69%).
• Classification of masses yielded an accuracy of 95.64% and an F1-score of 96.84%.

Conclusions:

• The integrated CAD system significantly outperforms conventional deep learning methods in all diagnostic stages.
• This system offers a promising tool to assist radiologists in the comprehensive analysis of breast masses from mammograms.