Closed espositoandrea closed 2 years ago
We've settled to an API consisting of two methods:
predict(image, format="txt|json")
: this method allows the upload a single image (a slice of an amyloid PET scan) and get the probability that the patient is affected by Alzheimer's disease.report(scan, format="txt|json|html|...")
: this method allows the upload of a NIFTI file (a PET scan) and get a report about the probability that the patient is affected by Alzheimer's, computed by applying the entire processing pipeline as well as the AI model. The format parameter determines the output formatting, and the accepted values will be determined more precisely down the line (for example, we may also introduce a way of generating a PDF report).I'll be defining the APIs in Python using FastAPI.
Although it is still in a pretty embryonic stage, the APIs have been defined and implemented, although the prediction does not use the model but rather a simple random function.
To provide the actual functionality using the AI model, I need the scripts and the model file that will be generated as part of #9, but I'll complete the work using the random functions and providing a simple way to execute a drop-in replacement of the classify
function.
We should propose a set of APIs to allow to use our AI model.