Purpose
The number of biopsy recommendations with B2 outcome especially according to the screening setting is critically discussed in the literature due to expensive overdiagnosis.
Study aimed to verify,
whether CAD-systems could potentially play a roleto reduce the number of unnecessary biopsies when applied at preinterventionally performed ML-images or mammogramsperformed for biopsy-planning.
Methods and materials
ML images of 350 histopathologically verified lesions,
sent for biopsy from the German screening programor from several gynecologists and radiologists performingcurative mammographies were evaluated using VuComp3.0.0.3(CAD1); R2Cenova1.3(CAD2) and iCADv.7.2(CAD3).
Size distribution of the lesions undergoing a biopsy was: microcalcification clusters: n=96<5mm; n=102<10mm;n=44<15mm;n=17<20mm;n=18<30mm;n=20>30mm,
masses and architectural distorsions,
respectively: n=21<10mm; n=15<15mm; N=19<20mm; n=15<30mm; n=17>30mm.
132 lesions were proven premalignant or malignant (37.7%),
whereas the remaining were scored as B3 or B2.
Selection for biopsy was done in a consensus conference ofseveral board certifiedradiologists and gynecologists.
Results
ML-images were processed in 158 cases (CAD1); 327 cases (CAD 2 and all 350 cases (CAD3).
Outcome of CAD 1 was: Sens:82.5%; Spec: 26.3; PPV: 42.6%; NPV:69.4%; outcome of CAD 2 was: Sens:91.8%; Spec: 13.2%; PPV: 38.6%; NPV: 73.0%; outcome of CAD 3 was: Sens:90.9%; Spec.
26.6%; PPV: 42.9%; NPV:82.9%.
Taking into account only screening-selected cases,
PPV of CAD 1,
2 and 3 was 26.3%; 31.5% and 32.9%,
whereas at curative cases this value was significantly higher: 69.6%; 46.3% and 53.7%.
Detection rate varied depending...
Conclusion
Only CAD3 supported the processing of all ML-images.
CAD-performance differs among available solutions significantly with most reliable and most specific outcome for CAD3,
whereas CAD2 was most sensitive and CAD 1 with the highest PPV-score at curative cases.
The diagnostic quality is,
however,
still not sufficiently reliable to allow a rejection of recommended biopsies according to a negative CAD analysis.
Personal information
It would be a pleasure to get in contact with you.
Thynk you for your interest.
Please contact me as follows:
Prof.
Dr.
med.
habil Ansgar Malich
Dr.-R.-Koch-Str.
39
99734 Nordhausen
Institute of Radiology
Sueharz-HOspital Nordhausen
Germany
[email protected]
fon: ++49 3631412858
References
Please send me an e-mail when asking further details.