Diffusion [MRI] histology (D-Histo) technique is very helpful in differentiating benign and cancerous prostate tissues. It can capture major differences between pathologies of these two types of tissues. This may be useful in the accurate detection of PCa.
However, it cannot help with differentiating prostate cancers of different risk levels very well. PCa with different Gleason scores are hard to be separated based on the pathologies captured using D-Histo techniques. Quality of data is a big drawback in this project, as well as the intrinsic difficulty of this classification task.
From the classification results, we can conclude that the model requires either more good data for training or a better set of features to learn the differences among the labels from. With higher-quality data, the classification task of PCa risk-stratification may become more feasible.