Figure 1.

Image processing workflow for Multi-Spectral Neural Network classification. DWI MRI image series from baseline and follow up scans are imported into the Clinical Trials Processor for de-identification and then registered using the normalized entropy mutual information method described previously⁷. Post registration, maximum and minimum intensities were computed, images intensity normalized over the interval [0.0, 1.0], and images were segmented using the MSNN method. Object maps were then individually inspected for classification errors, cleaned (if required), and then submitted to volumetric and IVIM diffusion modeling.