Radiologists outperformed a convolutional neural network (CNN) and radiomic analysis (RA) at classifying contrast-enhancing lesions on multiparametric breast MRI, according to a Nov. 13 study published in Radiology. With more training, however, CNNs may soon close that gap.

Daniel Truhn, with RWTH Aachen University in Aachen, Germany, and colleagues evaluated 447 patients with 1,294 enhancing lesions (787 malignant and 507 benign) from August 2011 through 2015. The team measured two RA approaches and a CNN algorithm against the interpretations of three breast radiologists.

Overall, radiologist interpretation achieved an area under the receiver operating curve (AUC) score of 0.98, beating both RA models (AUC, 0.81 and 0.78) and the CNN (0.88).

“A convolutional neural network was superior to radiomic analysis for classification of enhancing lesions as benign or malignant at multiparametric breast MRI,” Truhn, et al. wrote. “Both approaches were inferior to radiologists’ performance; however, more training data will further improve performance of convolutional neural network, but not that of radiomics algorithms.”

Truhn and colleagues found the performance of each RA approach did not improve when trained on a larger data set, while the AUC of the CNN algorithm jumped from 0.83 to 0.88 after doubling the size of the data set. Therefore, they noted, CNNs “might be able to mimic the elusive and subconscious process that occurs when a radiologist interprets MR images.”