 This study proposes a two-class GS classification methodology using a three-dimensional convolutional neural network with semantic segmentation to predict GS noninvasively using multi-parametric magnetic resonance images, MRIs. The results showed that the GS greater than or equal to seven precision and GS7 recall were significantly higher when using semantic segmentation compared to the normal volume. Additionally, the Ourockin segmentation volume was higher than that in normal volume. This article was authored by Takaki Yoshimura, Keisuke Manabe, and Hiroyuki Sujimori.