Image+metadata pre-trained model

In this release we provide:

1. A pre-trained model for prostate MRI scan classification .keras format, (model.zip)
2. The same model in onnx format, and (model.onnx)
3. A scaling factorcs csv needed for preprocessing the metadata (scaling_factors_df.csv)

The model performs classification of:

• 0: "T1-weighted axial",
• 1: "T2-weighted axial",
• 2: "diffusion-weighted images (DWI)",
• 3: "apparent diffusion coefficient (ADC)}

The CNN is trained using a combination of images and DICOM metadata from parts of the following collections in Imaging Data Commons (IDC):

1. QIN-Prostate-Repeatability
2. ProstateX
3. Prostate3T
4. ProstateDiagnosis
5. ProstateMRI
6. Prostate-fused-MRI-Pathology
7. Prostate-MRI-US-Biopsy