Heike Egger MD, Presenter: Nothing to Disclose

Christian Weismann, Abstract Co-Author: Nothing to Disclose

Christian Mayr, Abstract Co-Author: Nothing to Disclose

Martin Becker, Abstract Co-Author: Nothing to Disclose

Alexander Schlattau MD, Abstract Co-Author: Nothing to Disclose

Gabriele Konstantiniuk MD, Abstract Co-Author: Nothing to Disclose

The question is whether computer assisted analysis of ultrasound detected breast lesions is able to reduce the investigator influence on the number of true positive and true negative results. The calculated results are compared with the results of a well trained radiologist.

Computer aided diagnosis in ultrasound detected breast lesions B-CAD™ 2.3.1 (Medipattern Corp., Canada) is a software which is able to analyse a lesion by shape, orientation, margin, lesion boundary, posterior features, echo pattern and calcifications. In a retrospective study 150 histological confirmed breast lesions are investigated with a high frequency broadband ultrasound transducer, 6-16 MHz (Voluson E8, GE-Kretztechnik, Austria). The lesion size varies from 0.4 – 3.5 cm. 87 lesions are malignant 63 lesions are benign. The JPG-images of the lesions are analysed by B-CAD™ and by a well trained radiologist. After indicating the ROI to the B-CAD™-software the computer offers automatically up to 6 segmentation-proposals. The most appropriate segmentation is activated by the examiner and within seconds the software gives the result using the BI-RADS classification. Basically two decisions are offered by the software: BI-RADS 2 / BI-RADS 3 (no biopsy proposal) or BI-RADS 4 / BI-RADS 5 (biopsy proposal). This result is compared with the result of a well trained radiologist, who made his decision (biopsy / no biopsy) on the identical JPG-image of the lesion.

B-CAD™ shows a sensitivity of 98% and a specificity of 31%. The PPV is 66% and the NPV 93%. The accuracy is 70%. The results for the well trained radiologist are: sensitivity 97%, specificity 79%, PPV 90%, NPV 93%, accuracy 90%.  

On the level of this limited study the tested B-CAD™ 2.3.1 software provides high sensitivity and low specificity. The tested version is not able to replace a well trained radiologist.

The target to support the less well trained breast ultrasound investigator with a computer aided analyses by B-CAD™ 2.3.1 seems not to be achieved.

Egger, H, Weismann, C, Mayr, C, Becker, M, Schlattau, A, Konstantiniuk, G, Can B-CAD™ Help to Improve the Diagnostic Performance in Ultrasound-detected Breast Lesions?.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9001797.html