Open alif-munim opened 1 week ago
Yes, there are two ways to do this:
1) Modify the function predict metrics. Predict metrics essentially generates 50 different reports, and averages predictions across them. You would need to extend it to calculate appearances of "trivial","mild","moderate","severe" and then determine the optimal thresholds.
2) In case you don't want to deal with thresholds. You can fit a Logistic Regression for the categorical task at hand ("trivial","mild","moderate","severe") on top of echo prime embeddings.
Let me know if this answers your question and I will close the issue.
Hello, we were hoping to use EchoPrime to assess conditions such as dilation, stenosis, and regurgitation in echo studies based on severity (e.g. mild, moderate, severe). Is this something that is currently possible?