Computer-assisted Diagnosis in Full-field Digital Mammography-Results in Dependence of Readers Experiences

Abstract

The object of this study was to assess the clinical usefulness of computer-assisted diagnosis (CAD) in the interpretation of early-research, benign and malignant mammograms in dependence on readers' experience with CAD. CAD was applied on digital mammograms of 303 patients who were divided into three groups: early-research (n=103), benign (n=102) and malignant group (n=98). Mammograms were analyzed by three readers with different experience in evaluating mammograms (medical student, an assistant and an attending physician specifically trained in mammography). All images were presented accidentally with and without the influence of CAD and from different patient groups. The mammograms were classified according to BI-RADS classification. To evaluate readers' sensitivity and specificity with and without the application of the CAD system, ROC analysis and the corresponding area under the curve (AUC) were evaluated for each reader. Afterwards significant differences of the accuracy according to readers experience and according to the assistance of the CAD system were calculated. All readers have an account of accuracy by using CAD in both patient groups. The highest benefit has the student (10% increase of the AUC) followed by the resident (4%) and at least followed by the mammography fellow (3%). There are significant varieties of the accuracy in addiction to the readers' experience and to the examination method with and without CAD system. Patient group has not a significant influence to the elevation of accuracy by using the CAD. All three readers have nearly the same increase of AUC in the examinations of malignant and early-research group summarized and of the malignant group only. Finally, the increase of accuracy depends on the readers' experience. For all patient groups CAD-application causes a steeply increase of the ROC curve and consequently a gain of sensitivity.