Image Processing

I built a Python and Streamlit interface around a NIfTI preprocessing workflow, bringing visualization, denoising, registration, segmentation, edge detection, and skull-removal controls into one place. I have not re-verified the processing pipeline itself.

Three abstract imaging volumes connected through a staged transformation pipeline

Problem

Running every preprocessing operation as a separate script made the workflow hard to explore, so I wanted each step and its visual feedback in one place.

Constraints

I had to work with large, specialized NIfTI data
I treated every medical image as sensitive by default
I needed each processing step to have understandable controls and a visible preview
I kept patient and clinical files out of this case study

Technical decisions

I built the interface in Python
I used Streamlit so I could iterate quickly on controls and previews
I organized the interface around separate preprocessing stages
I put a visual preview beside each set of controls so the intended transformation stayed understandable

Evidence

I still have the public Python source for the Streamlit interface; processing execution remains unverified
I did not copy medical datasets, patient files, or clinical records into this public case study
I could not re-verify processing execution because the hosted app remained on its loading screen after I requested a wake-up

What I learned

What stayed with me from this project is the value of making technical transformations observable while treating medical data as sensitive from the start.