Significance testing of machine learning features

Hi everyone:

I used the MRI features to predict disease severity with an R-squared value around 0.015, using the SHAP method to provide the importance of the MRI features.

However, SHAP only gave me ranking (or importance of featrues), which prevented me from focusing and doing the next analysis. I would like to know how to get the result of significance (which MRI feature successfully predicts symptoms). Do you think there is any good way to do it? I would be eternally grateful!

enter image description here