Steepest changes of a probability-based cost function for delineation of mammographic masses: a validation study

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Radiology

Background:

• Accurate segmentation of mammographic masses is crucial for early breast cancer detection.
• Existing methods for mass boundary detection require manual input and can be time-consuming.
• Developing automated, reliable segmentation techniques is a key goal in medical image analysis.

Purpose of the Study:

• To develop and rigorously test an automatic boundary detection method for mammographic masses.
• To evaluate the performance of a steepest change analysis technique for mass segmentation.
• To statistically validate the developed method against expert annotations.

Main Methods:

• A novel segmentation method utilizing steepest change analysis on a composed probability density cost function was developed.
• The method was applied to 124 digitized mammograms from the Digital Database for Screening Mammography (DDSM).
• Segmentation results were validated using overlap, accuracy, sensitivity, and specificity metrics against manual traces from two expert radiologists.

Main Results:

• The developed method successfully delineated mammographic mass boundaries, with steepest changes in the cost function identifying optimal delineations.
• Statistical analysis, including Analysis of Variance (ANOVA), confirmed the significance of the findings (p-values as low as 1.03 x 10(-2)).
• The automated segmentation results showed a closer correlation with one expert radiologist compared to the second.

Conclusions:

• The steepest change analysis technique provides an effective automatic boundary detection method for mammographic masses.
• The optimal intensity threshold for segmentation corresponds to a specific steepest change location in the probability density function.
• The method demonstrates potential for improving the efficiency and consistency of mammographic mass analysis in clinical settings.