A new information-theoretical model for the performance evaluation in breast cancer imagery

The results of using X-ray mammography and ultrasound as well as other imaging modalities for breast cancer screening vary widely. New methods are needed for a more consistent assessment of the value of screening and for a more reproducible cost-benefit analysis. A new information-theoretical approach is presented herein in order to address these needs. In this approach, the entropy calculated from diagnostic and predictive data is used to measure the amount of clinical information provided by the different imaging modalities. In order to more consistently compare different studies, the concept of information efficiency is introduced. It is further shown herein how value and cost can be assigned to the provided information, such that a return-on-investment ratio, as a global measure of performance, can be calculated. The proposed model is evaluated for breast cancer screening.