See NeuroCog.ai in Action

A short talk on the science behind NeuroCog.ai

Transcript

Lightly edited for readability. Asides about the room and the slides have been removed, and one self-correction resolved to what the speaker settled on; the substance is unchanged.

Our next speaker is Dr. Walter Sena, an IMH Marshall Fellow, supervised by Dr. Todd Woodward.

Hello. I'll talk about what I've been working on these past years: a neurocognitive assessment spanning all DSM-5 domains, designed for low-frequency eye trackers.

This is Jack's curve, which is well known in dementia research. This particular example is illustrative rather than based on real data; it is here so we can follow the concept. Here is a CSF collection for detecting biomarkers, and a PET scan, in this case an amyloid PET scan. Both come before the other markers: blood-based biomarkers, neurocognitive assessment, and finally daily function. The point is that the more invasive biomarkers turn positive first, and that is the problem.

The solution is this: a neurocognitive assessment based on an eye tracker. People understand it better if I call it a gaze tracker. Here is how it works. We know where someone is looking while they take the test, and this is what sits behind that: fixations and saccades. At the bottom you can also see moment-to-moment measures of what is conventionally called attention, though I think naming cognitive domains is somewhat arbitrary. That is another talk. Notice that longer saccades show up differently in the analysis below. There is a great deal we can recover from gaze tracking. These are memory tests where we would normally report accuracy alone, and here we have considerably more.

So what is the relevance? The idea is to combine this with blood-based biomarkers, which are far less invasive, and there are many of them. In Alzheimer's disease we have been talking a great deal about GFAP, neurofilaments, the Aβ42/Aβ40 ratio, and phosphorylated tau. We have those, and more, for other diseases as well, alongside a battery of cognitive tests. We are talking about at least 30 cognitive tests, each with roughly 10 outcomes.

From that we can estimate the risk of developing particular neurocognitive disorders, including Alzheimer's disease, and not only Alzheimer's but its subtypes. In fact, rather than saying Alzheimer's disease or frontotemporal dementia, it is better to describe the finding as amyloid beta accumulation or phosphorylated tau accumulation. More precisely still, we should specify amyloid beta accumulation in the prefrontal cortex, or in the hippocampus, as these are less biased ways of referring to it.

We might then generate a medium-risk score, meaning that the person does unfortunately need to proceed to invasive testing; or a high-risk score, in which case treatment should begin immediately. That would be lecanemab, which is available now, or donanemab, from Eli Lilly, which has produced good results.

Why does this matter? If we delayed Alzheimer's disease by one year, we could reduce future prevalence by 11% and cut annual care costs in the United States alone by 219 billion dollars. That has been published.

The assessment itself takes only five minutes, and you get your results at the end. They are raw results for now, of course, since we still need to collect data to normalise them. That's it. Thank you.

INVESTORS MATERIAL