HMS-CardiacMR/Radiomics_Histology_NIDCM — reverse-engineered prompt

Build me a clean research workflow for studying whether cardiac MRI radiomics are associated with biopsy histology in patients with non ischemic dilated cardiomyopathy.

I want to start with MRI image and mask data stored in mat files, where one matrix is the image, one is the mask, and pixel spacing is saved as px_size. The workflow should extract radiomic features for each MRI sequence using pyradiomics, but leave out shape features. Then it should remove highly correlated features, help choose the number of feature clusters with consensus clustering, create hierarchical clusters, pick one representative medoid feature from each cluster, and run logistic regression models against histology findings.

Please make it easy to run step by step, with clear input and output folders, readable saved tables, and plots where useful. The final results should include odds ratios, 95 percent confidence intervals, and enough documentation that a researcher can rerun the analysis on their own data. Look up current docs online if you need to.