Cerebral Microbleeds

Cerebral Microbleeds with 3D Cascaded Network

Steps to execute code

Groundtruth and data generation

• Run adni_dcm2nii.py file to convert all DCM (DICOM) files to Nifti File format
• Run adni_groundtruth_generation.py file to get ground truth from CSV file, modify the condition (if folder == 'ADNI 4': ) to get ground truth for entire data

Normalization

• Normalize the actual images with Normalization.py file

Generating input patches

• Run Generate_input_patches.py to get patches from ground truth and the actual image
• Change ( img_dir ) and (out_dir) for generation of patches.
• Two directories 'in_patches_16_16_10' and 'out_patches_16_16_10' are present as locations to save the patches.

dataset preparation

• Run balanced_dataset.py and complete_dataset.py to get pickled files(.sav)