Developing a utility decision framework to evaluate predictive models in breast cancer risk estimation.

Area of Science:

• Oncology
• Radiology
• Genetics

Background:

• Radiogenomics integrates imaging and genetic data for disease prediction.
• Established methodologies for evaluating radiogenomics predictive models are lacking.
• Breast cancer diagnosis and progression prediction are critical clinical challenges.

Purpose of the Study:

• To develop a decision framework for assessing radiogenomics predictive models in breast cancer diagnosis.
• To evaluate the utility of combining Gail risk factors, mammographic features, and single nucleotide polymorphisms (SNPs).
• To establish optimal operating points for maximum expected utility in breast cancer risk estimation.

Main Methods:

• A retrospective case-control study collected Gail risk factors, SNPs, and mammographic features.
• Three logistic regression models were constructed using different feature combinations: Gail, Gail + Mammography, and Gail + Mammography + SNP.
• Receiver operating characteristic (ROC) curves were generated, and utility analysis was applied to determine optimal operating points.

Main Results:

• The model incorporating Gail factors, mammographic features, and SNPs demonstrated superior predictive performance.
• Utility analysis identified optimal operating points on ROC curves for maximum diagnostic utility.
• McNemar's test confirmed that SNPs and mammographic features significantly improve breast cancer risk estimation.

Conclusions:

• The developed decision framework, incorporating utility analysis and statistical testing, effectively evaluates radiogenomics predictive models.
• Integrating mammographic features and SNPs significantly enhances the accuracy of breast cancer risk prediction.
• This approach provides a robust method for assessing the clinical utility of radiogenomics in breast cancer diagnosis.