Model selection framework should be an in-built feature.

[Devguru-codes]

There are a lot of models present like for DCE-MRI, the choice is between Tofts, Extended Tofts, Patlak, and 2CXM. There is no way to fit these models altogether and compare their results and then choose the best one, and the R² metric we use prefers complex models (which might be overfitted). I am thinking to create a class that can take a list of models and fits all of them to the same data and compare results and suggest the best model. The metric we can use in it could be discussed. I think use of akaike information criterion and bayesian information criterion that penalises complex models keeping overfitting in check while R² tells us how well the model actually fit. to concise it, create a best model selection framework in which user passes model names and then we will fit the model and give the best model after fitting on given data to the user. @ltorres6 sir, lets discuss on this feature and then I will work on it.

[ltorres6]

Thanks for the suggestion. This methodology is typically not used for perfusion fitting. The model is selected based on the expected pathology or physiology, and statistical model selection using AIC and BIC is not usually performed as it is in other statistical analyses.

[Devguru-codes]

Thanks for the explaination @ltorres6 sir.