How a Neurologist Is Shaping the Next Generation of Medical AI

By Dr. Paul Fletcher — 2025-10-25

Dr. Paul Fletcher trained in neurology because he was drawn to complexity — the organ that generates consciousness, language, movement, and personality, and the spectacular ways it can break down. Thirteen years into his consultancy at a major university teaching hospital, that same love of complexity drew him to AI evaluation work.

Why neurology is uniquely difficult for AI

"Neurology sits at an intersection that AI finds genuinely difficult," he explains. "The symptoms are often vague and overlapping. The diagnostic process requires integrating history, examination, imaging, and electrophysiology in ways that don't reduce to pattern matching. And the stakes are extremely high — a missed stroke, a delayed MS diagnosis, an unrecognized epilepsy syndrome. These have life-altering consequences."

Fletcher's IXO work focuses primarily on diagnostic reasoning evaluation — testing how neurological AI performs when presented with ambiguous presentations, rare conditions, and cases where the correct answer requires synthesizing information across multiple modalities.

The stroke mimic problem

One category he's found particularly important is AI performance on stroke mimics — conditions that resemble strokes but require entirely different treatment. "An AI trained mainly on confirmed stroke cases will over-diagnose stroke in anything that looks like it. That's dangerous. Training it to recognize migraine with aura, hypoglycemia, Todd's paresis — that requires a neurologist who's seen those cases and knows what differentiates them."

He contributes roughly eight hours per week to IXO projects, mostly on weekend mornings. "The work is genuinely interesting. It's not rote labeling — it requires real clinical thought. And the compensation is honest about what specialist knowledge is worth."

The bigger picture

"Medicine is going to use AI. The question is whether it's AI that neurologists helped build, or AI that was built without us and then handed to us. I'd rather be in the room."

Read How a Neurologist Is Shaping the Next Generation of Medical AI on the IXO blog