I asked for my MRI data. Then I built a way to explore it.

A couple of weeks felt too long to sit with that uncertainty.

In 2023, I suddenly lost all hearing. Doctors told me it was likely to be a tumor in my brain and that an MRI would be needed. By the time I got the scan, I was anxious to know what was happening.

Then I was told it would take a couple of weeks to analyze the MRI and return my results. That wait was hard to accept. I asked for a copy of my raw MRI data so I could start exploring the information I already had.

I used AI to help me build a tool that reconstructed a three-dimensional view of my brain. Python handled the image-processing work, and Blender gave me a way to move through the result. I had not seen this particular approach before; I had a need, access to the data, and a reason to try.

Start with the slices. Build the visual continuity between them.

I built a Python tool that read the brain-image slices from disk and prepared them for a three-dimensional workflow. The input was the scan material I already had. The aim was to make those separate images useful together inside a creative 3D environment.

The gaps between slices were an important part of the problem. Simply placing the original images next to one another would leave the viewing experience limited by the separation between them. I used generative AI to interpolate intermediate images and add visual continuity across those gaps.

Those intermediate images extended the sequence that would become the Blender stack. They were estimates between the original slices, giving the presentation more steps to move through. That distinction is part of the method: the scan images supplied the starting information, while the interpolation supplied the additional visual transitions.

This joined data handling with image generation in one workflow. Reading the files was the entry point; preparing a sequence that could be assembled and explored was the actual product of this stage.

Empty image space should not block the view.

Stacking image rectangles is not enough to create a useful internal view. The background around the scan content can become a series of opaque surfaces, hiding the structure behind it. Before assembling the images, I needed to deal with that negative space.

I used image processing to remove the backgrounds and turn them into transparency. The intent was to carry the visible scan content into the stack while allowing the empty parts of each image to disappear from the presentation.

That changed what the stack could communicate. Instead of encountering a succession of solid image backgrounds, the viewer could look through the transparent regions and follow the forms across layers. Background removal was therefore part of the spatial design, not just a cosmetic cleanup step.

I could move through the data and examine it from different angles.

The workflow brought the original slices and generated intermediate images into Blender as a stack. With the image backgrounds made transparent, I could move through the structure and look at how the layers related to one another. The demonstration shows multiple viewpoints and colored scan structures.

I am no brain surgeon. I did extensive research to understand a little of what I was looking for. To my untrained eye, the structures looked symmetrical, and I did not see an obvious abnormality around the areas connecting my brain to my ear.

My own, unprofessional impression was that there might not be a tumor. That helped me sleep better while I waited. It was not a diagnosis or a way to rule a tumor out; the official interpretation was still necessary. The generated intermediate images were visual estimates, not additional MRI measurements.

Move from a layered view toward a modeled volume.

I later added a voxel-based algorithm to create an actual three-dimensional model. This extended the work beyond the initial arrangement of image slices and interpolations. The project now had two related forms: a navigable image stack and a model built through a volumetric representation.

That progression mattered to the direction of the tool. The stacked view made the scan sequence tangible, while the later modeling step explored how the information could become a three-dimensional object in the workflow.

I see those as successive stages of the same experiment. First, make the images accessible as space. Then develop that spatial representation into a model. Keeping the purpose of each stage clear made the overall project easier to explain.

My family and I celebrated the result. I also celebrated the build.

When the official results came back, they confirmed that I was tumor-free. My family and I celebrated that news.

I was also celebrating what I had managed to build. I had set out to do something I had not seen done before, made it work, and used it to reach my own preliminary understanding before the report arrived. Building gave me something useful during the wait; the medical report provided the confirmation.

I later made the original tool available as a free download and published the Blender demonstration on November 29, 2024 through the keiazo and Asaday Instagram accounts. A personal need had become a project other people could explore.

That is the same thinking I applied to AfroTech.guide. It began with about 12 people trying to coordinate their conference week. In both cases, I started with a need I understood and built something I wished I had. I did not wait for perfect conditions to begin.

Watch the original November 2024 demonstration →

Read how AfroTech.guide started with about 12 people →