Breast MRI Lesion Classification: Comparison between Human Readers and Backpropagation Neural Network (BNN)
Arbash Meinel L, Stolpen A, Berbaum K, Fajardo L, Reinhardt J
Philips Research USA
Five radiologists with different levels of experience manually classified eighty breast MRI lesions both with and without CAD assistance. The performance of the computer classifier and human readers were compared using ROC curves, which showed that human readers significantly improved with assistance (p < 0.05), and the human reader? performance was also evaluated using MRMC analysis, which showed that the human reader performance with and without CAD system assistance can be generalized to the population of cases with a statistically significant improvement (p < 0.05).These results show significant advantages to using CAD in classifying BMRI lesions.