Would you take a test that could predict the potential onset of dementia? An artificial intelligence (AI) tool from Cognetivity Neurosciences is allowing patients to do just that. Through either its iPhone app OptiMind or Cognica , a tool designed explicitly for use in medical settings in conjunction with a care team, patients can use Cognetivity’s test to detect the presence of even a mild cognitive impairment.
Mild cognitive impairment can emerge long before any significant symptoms of dementia. Many patients may not notice any changes at all, but this window is a crucial one wherein interventions can be especially effective in prolonging a patient’s cognitive function.
The test from Cognetivity is an image categorisation test designed to engage specific areas of the brain affected in pre-symptomatic stages of cognitive decline. Users are shown a series of natural images and must classify them as being either animal or non-animal pictures. The software detects subtle impairments in information processing speed to try and catch formative signs of the disease before it starts to have a significant impact on a patient’s life.
Over the past year, the company has gone from strength to strength, and recently received approval from the US Food and Drug Administration (FDA) to market CognICA as a medical device. The test has been deployed in clinics across North America, the UK and the UAE, in collaboration with a number of leading healthcare providers. Cognetivity has partnered with Loveday & Co., the only dedicated operator of specialist memory care in the UK, teamed up with a number of NHS Trusts, and recently reached a commercial agreement to deploy the tool at the Clemenceau Medical Center in Dubai.
Outside of memory care, the tool has been rolled out across the US and Canada in the clinics of Ketamine One, a health and wellness company that aims to treat mental health conditions through ketamine-assisted therapy and psychedelic medicines.
Medical Device Network sat down with Cognetivity chief financial officer Tom Sawyer to learn more about the company’s recent deployments and the challenges of regulating AI.
Chloe Kent: Where is Cognetivity at as a business right now?
Tom Sawyer: Cognetivity is in a very exciting place. We’ve been through a number of years of development of the technology and we really feel we’ve de-risked the science. We’ve published peer-reviewed papers, we’ve de-risked the regulatory process with our CE marking and our FDA registration. We’ve built a fantastic technology architecture with what we believe is a very, very good product. It’s very usable, very adaptable, very integratable. It’s a lovely place to be as a company.
Now our challenge is to get it out there and make it available to as many people as possible.
CK: Why is early intervention so important when it comes to conditions like dementia?
TS: It’s extremely important to detect early. Even in the absence of a disease-modifying therapeutic in the UK, although one has now been approved in the US and UAE, early detection makes a huge difference because there are many things you can do. If you learn about it early, before becoming severely or even moderately impaired, you can prolong your functional life for considerable amounts of time.
It’s fairly common-sense stuff. Things like sleeping better, stopping drinking and smoking, better diet and exercise. These are known as modifiable risk factors. If you improve them, you massively improve the prospects for your brain health.
Obviously that has great benefits to the patient, but also for payers. Healthcare systems benefit hugely from early detection because it greatly reduces the lifetime cost of care for patients. That’s very important, because with increasing strain on health services all over the world, the idea that early detection can reduce financial and capacity pressure is critical.
CK: Where is Cognetivity’s technology currently in use? Can you talk about its applications outside of memory care?
TS: The clinical tool is not something that is available to the public, for the very reason that it’s designed to be used by doctors in context of other clinical decisions. But in terms of the Optimind platform, that’s freely available on the App Store.
We have a few early customers, including a handful of NHS Trusts in England. We also have it in use in in elderly care and in psychedelics clinics in North America. We’ve partnered with a firm called Ketamine One, which is tackling treatment-resistant depression and post-traumatic stress disorder (PTSD). A loss in cognitive function is very noticeable and measurable in depressed patients and one of the signs of that improving can be cognitive function.
Our customers are all looking for a practical, objective tool that can fit into their clinical pathway, which enables them to get a measure of the response to treatment of their other patients.
CK: How does AI fit into this picture? How do your solutions actually work?
TS: The engine that drives CognICA and Optimind is exactly the same. You have a short duration of exposure to natural images and you react to whether or not you’ve seen an animal. The experience is intrinsically gamified, rather than feeling like you’re undergoing an examination. Someone sitting with a clipboard and asking you questions that are going to have an impact on how your health is perceived is very stressful. It’s far less stressful to sit with an iPad in front of you and play a game where you’re just reacting to things. It’s much more passive, much more enjoyable and much faster.
The AI then allows us to compare somebody’s pattern of responses to known data that we’ve collected under stringent clinical conditions. The AI is then able to say to what extent the results look like these previously recorded datasets.
What that gives you is a very important mathematical edge. Traditional pen and paper tests are on a linear scale; a difference of plus or minus one point can change the category you fall into, so they’re very blunt instruments. The AI allows you to look at data in a multidimensional fashion so it means the cut-offs are much more dynamic, reducing the risk of false positives and false negatives.
CK: Regulating AI can be difficult since it’s inherently changeable. How do you make sure that your product is still the thing that’s been approved?
TS: What we do with regulators is approve a version of the AI. Once we submit it for regulatory approval that is then locked, so effectively that’s like the FDA version of the software. That will then stay exactly the same until such time as we believe we have a better performing model or a model that does something different, which we would them submit separately.
From a regulatory point of view, it’s very important that the version that has been approved remains the same, and regulators have been very correctly insistent upon that.