Computerized assessment of tissue composition on digitized mammograms

Area of Science:

• Radiology
• Medical Imaging
• Biomedical Engineering

Background:

• Accurate breast tissue composition assessment is crucial for mammography interpretation.
• Subjective radiologist ratings can introduce variability.
• Objective, quantitative methods are needed to enhance diagnostic consistency.

Purpose of the Study:

• To develop and evaluate a computerized method for quantitative assessment of breast tissue composition.
• To compare the computerized method's accuracy against expert radiologist consensus.
• To establish a reliable, objective measure for breast density on mammograms.

Main Methods:

• Digitized mammograms (n=200) were reviewed by three radiologists for tissue composition (0-4 scale).
• A consensus rating was established as the reference standard.
• A computerized method was developed involving tissue segmentation and feature computation.
• A summary index was created to replicate radiologist ratings, with Pearson correlation used for evaluation.

Main Results:

• Individual quantitative features showed high correlation (r > 0.8) with subjective radiologist ratings for dense areas.
• The developed computerized summary index demonstrated significant correlation (r = 0.87) with the consensus ratings.
• The method successfully reproduced radiologists' assessments of breast tissue composition.

Conclusions:

• Computerized methods can effectively quantify breast tissue composition from mammograms.
• The developed method shows strong agreement with expert human interpretation.
• This technology has the potential to standardize breast density assessment in clinical practice.