Data analysis for detection and localization of multiple abnormalities with application to mammography

Rationale and Objectives. In assessing diagnostic accuracy it is often esse ntial to determine the reader's ability both to detect and to correctly loc ate multiple abnormalities per patient. The authors developed a new approac h for the detection and localization of multiple abnormalities and compared it with other approaches. Materials and Methods. The new approach involves partitioning the image int o multiple regions of interest (ROIs). The reader assigns a confidence scor e to each ROI. Statistical methods for clustered data are used to assess an d compare reader accuracy. The authors applied this new method to a reader- performance study of conventional film images and digitized images used to detect and locate malignant breast cancer lesions. Results. The ROI-based approach, the free-response receiver operating chara cteristic (FROC) curve, and the patient-based approach handle the estimatio n of the false-positive rate (FPR) quite differently. These differences aff ect the measures of the respective areas under the curves. In the ROI-based approach the denominator is the number of ROIs without a malignant lesion. In the FROG approach the average number of false-positive findings per pat ient is plotted on the x axis of the curve. In contrast, the patient-based approach mishandles the FPR by ignoring multiple detection and/or localizat ion errors in the same patient. The FROG approach does not lend itself easi ly to statistical evaluations. Conclusion. The ROI-based approach appropriately captures both the detectio n and localization tasks. The interpretation of the ROI-based accuracy meas ures is simple and clinically relevant. There are statistical methods for e stimating and comparing ROI-based estimates of accuracy.