Abstract #2889

Su M, Yu H, Chen J, Chu Y, Nie K, Nalcioglu O

33 patients who had a baseline and a F/U MRI study after 1-2 cycles AC were included (17 responders and 16 non-responders). An artificial neuronal network analysis based on the baseline morphology/texture/kinetics parameters was performed to find optimized subset for response classification. A quantitative analysis was performed to characterize the morphology and texture features (10 GCLM and 14 Laws). By adding the enhancement kinetic parameters, the classification accuracy improves from 76% to 85%. We demonstrated that quantitative analysis of morphology and texture in breast cancer is feasible, and artificial neuronal network may be applied to find an optimized subset for diagnosis, or therapy response prediction.