Related Experiment Video
Updated: Jun 26, 2026
10:33
Correlating Behavioral Responses to fMRI Signals from Human Prefrontal Cortex: Examining Cognitive Processes Using Task Analysis
Published on: June 20, 2012
Correlative feature analysis on FFDM.
Yading Yuan1, Maryellen L Giger, Hui Li
1Department of Radiology, Committee on Medical Physics, The University of Chicago, Chicago, Illinois 60637, USA. yading@uchicago.edu
Medical Physics
|January 30, 2009
Summary
This study developed a computer framework to match mammogram lesion images from different views. The system accurately identifies corresponding lesion images, improving diagnostic tools.
Area of Science:
- Medical Imaging
- Radiology
- Computer-Aided Diagnosis
Background:
- Matching lesion images across mammogram views is crucial for diagnosis.
- Breast nonrigidity and mammogram projection complicate lesion correspondence.
Purpose of the Study:
- To develop a computerized framework for differentiating corresponding lesion images from noncorresponding ones.
- To enhance diagnostic accuracy in mammography using automated image analysis.
Main Methods:
- A dual-stage segmentation (RGI and active contour) extracts mass lesions.
- Lesion features (density, size, texture, neighborhood, distance to nipple) are automatically extracted.
- A two-step Bayesian artificial neural network (BANN) scheme calculates correspondence probability.
Main Results:
- The distance feature achieved an AUC of 0.81 +/- 0.02.
- A feature subset (distance, gradient texture, ROI correlation) yielded an AUC of 0.87 +/- 0.02.
- Multi-feature analysis significantly outperformed single features.
Conclusions:
- The proposed framework effectively distinguishes corresponding mammogram lesion pairs.
- Automated feature extraction and BANNs improve lesion matching accuracy.
- This approach has potential to enhance both radiologist and CAD system performance.
