Why I Don’t Have a Direct Neuroimaging Startup

Yesterday I wrote about medical image processing. Today I write about a particular area of medical image processing dealing with the brain: neuroimaging.

MRI scans are good at visualizing internal soft tissues and are used to image the brain. The brain is composed of gray matter, white matter, and cerebral spinal fluid. Gray matter and white matter are made up of neurons. The neuron cell bodies are located in the gray matter towards the surface of the brain, while the axonal connections are located in the white matter towards the center of the brain. The axons are what connect different regions of the brain together, like wires communicating.

Neuroimaging can be broken into 3 main types:

Structural – this type of imaging just looks to see what the structural integrity of the brain organ is, e.g. are there deformities (for instance in head trauma situations), is there a tumor, is there a problem with the blood supply, etc.
Functional – this type of imaging attempts to see which areas of the brain are activated during specific activities, e.g. to do research to understand which parts of the brain are responsible for different behaviors (and to test drugs), to attempt diagnoses of mental health, motor skill deficiencies, or other ailments
Connectivity – this type of imaging is focused on trying to assess the connectivity of white matter brain regions. The goal is to figure out if the wires are working well or if they are breaking down, e.g. as they do in Alzheimer’s. This type of imaging is also highly experimental and helps answer questions about how the brain is wired.

My PhD research was focused on the 3rd type of neuroimaging. I’ll write about that more tomorrow.

While there is a lot we have learned about the brain, it is the least understood organ in our body. It is incredibly complicated and no one really understands the basic mechanisms of conscious thought and memory storage.

Neuroimaging software is largely dominated by the manufacturer’s of the MRI machines: GE, Siemens, and Phillips. It is overly hard for startups to break into that space given how tightly coupled the software is to those massive hardware purchases. This is the primary reason why I have not directly commercialized neuroimaging software as a startup.

ArrayFire does, however, address a lot of neuroimaging problems indirectly by providing computational tools to enable other neuroimaging software to run faster. While I was in school, there was a moment when I had to recognize that my direct area of research was not commercially viable and that I should find an adjacent space where I could better build something into a business.

Have you found that your direct areas of expertise are not commercially viable? Did you have to pivot to something more tractable?

Related articles