Microcalcification detection in full-field digital mammograms: A fully automated computer-aided system.

Area of Science:

• Medical Imaging
• Radiology
• Computational Pathology

Background:

• Microcalcification clusters in mammograms are early indicators of breast cancer.
• Detection is challenging due to breast composition, anatomy texture, microcalcification size, and low mammogram contrast.
• Automated tools are needed to support clinical diagnosis.

Purpose of the Study:

• To develop and evaluate an automated three-phase method for detecting clustered microcalcifications in digital mammograms.
• To provide a tool that assists clinicians in identifying early signs of breast cancer.

Main Methods:

• A three-phase approach involving pre-processing to highlight breast structures.
• Single microcalcification detection using Hough transform to identify characteristic shapes.
• Cluster identification using a clustering algorithm that incorporates expert rules.

Main Results:

• The method was evaluated on 364 mammograms from 182 patients.
• Achieved a true positive ratio of 91.78% with 2.87 false positives per image.
• Demonstrated high performance, even in challenging cases with dense breast tissue.

Conclusions:

• The proposed method effectively detects microcalcification clusters in mammograms.
• Performance is comparable to state-of-the-art methods without requiring training data.
• Maintains high accuracy in difficult clinical cases, including those with dense breast tissue.