Convolutional neural networks to predict brain tumor grades and Alzheimer’s disease with MR spectroscopic imaging data
Fig 6
Important spectral regions for the separation of brain tumor dataset grade II vs grade III.
The average spectrum for each area is plotted as a reference in (solid) black and the gray zone around it shows the standard deviation. The red dotted lines indicate the center of the most important regions (found by stability selection) and the red shaded areas around them indicate their size. These areas were obtained using stability feature selection on the output of the convolutional layer of the trained SL-CNN using only spectra as inputs.