New Model Data Preparation

[CavRiley]

If this feature request related to a problem? Please describe.

The dcm-classifier package lacks a script for preparing data for new model training. The prepared data should include a set of json files in the following format ${ImageType}_${SeriesUID}.json and the accompanying nifti file formatted like this {ImageType}_${SeriesUID}_${SubvolumeNumber:03}.nii.gz.

Alternatives Considered If Applicable

N/A

Rationale

New model training using image pixel data can provide features the current method of classification does not include potentially improving image classification.

Implementation Ideas

We can create the data needed for training by using the image metadata gathered in the current implementation of the model as well as the nifit image we create. The JSON file should contain the image type, metadata for classification as well as the output probabilities for each sub-volume within the series.

Additional Context

N/A