Let me just kind of go over a little bit about what I want to talk about. So I’m a neurologist. My specialty in neurology is actually in diseases of the nerve and muscle, something I’ve been doing for about 30 years now. And over the past decade or two in many areas of neurology and in medicine in general, there’s been this phrase that people use, which is finding the right drug for the right person. And I like this phrase because it uses the word person. It doesn’t use the word diagnosis. And the reason for that is a diagnosis is a very easy thing to make, as long as you’ve got a couple of initials after your name. So M-D-D-O-N-P-P-A, anybody can make a diagnosis. But that doesn’t mean that it’s the right diagnosis for that patient or that you’ve actually achieved the right disease for that patient. And unless we do that, we can’t then determine the right drugs for the patient. So let me give a couple of examples that are probably in the front of everybody’s minds right now. So we all know that the anti-amyloid drugs are now being approved for Alzheimer’s disease. We also know that if you go to the best dementia centers in the country and you then do autopsies on patients who’ve been diagnosed with Alzheimer’s disease, that about 30% of those patients don’t have Alzheimer’s disease when they die. They have a different form of dementia. And about another 10% to 15% actually have two forms of dementia, Alzheimer’s, let’s say, plus strokes. And so that means almost half of the patients that walk around with this diagnosis of Alzheimer’s disease probably don’t have it. So how do we then in a field that’s now going to be tens of billions of dollars affecting Medicare in a matter of months, how do we achieve a better way to come to that right diagnosis? And that’s true in multiple sclerosis. It’s true in autoimmune diseases across the board. So I’m going to talk about using an AI-assisted diagnostic algorithm in a peripheral nerve disease. Our technology really expands to any disease state that you want to have as long as you can define a set of criteria that define that diagnosis. So in this disease CIDP, which is an autoimmune disease where a person’s immune system starts to attack their nerves, their nerves get very weak. John Katz and I were involved in a very long project that we published about five years ago with other experts across the country and with the support of FFF. And what we found was if you took patients that were diagnosed by the doctor as having this disease, CIDP, who then went to the insurance company in the prior authorization process, and the insurance company medical reviewers said, yes, we believe it’s CIDP, we’re going to put these patients onto therapy that can range from $100,000 to $300,000 a year, that when you actually asked experts to review the cases, only about one in three actually might have the disease. And only about one in seven actually met the criteria that the insurance company laid out for themselves. So they had laid out the rules and said, we have to follow these rules. Yet their prior authorization process, their medical experts only followed that in one of seven of the cases. And so we published this a few years back. And about the same time, two other real experts in the field, Jeff Allen and Rich Lewis, took a different approach, which was to say, as neuromuscular experts, they were referred cases into their tertiary centers all of the time for the question of, well, I think this is CIDP coming from the primary doctor. What do they think it is? And they found that only half of the patients actually met the criteria, despite having been treated with these very expensive immunomodulatory therapies. And this is one of my favorites. So if you’re doing a clinical trial, let’s say in CIDP, and you’re the manufacturer of that drug doing your Phase II, Phase III trial, you have to go out and find the top experts in the world and ask those experts like the insurance companies do to review the cases. And then to decide, should that case get into the research trial or should it not get into the research trial? So these are the top experts in the world. And it turns out that their agreement with each other was basically the same as flipping a coin. It was 50% of the time. So even among experts, the top experts in the world, they can’t even agree if the case meets the definition or doesn’t meet the definition. So what did we learn from that study? What we learned is, if you give doctors hundreds of pages of records and you ask them to pile through these records and come up with a diagnosis, it’s pretty darn confusing. It takes a ton of time. And in a case of like CIDP, there are many different pieces of information that have to be integrated. I mean, probably, honestly, hundreds of pieces of information. So if that reviewer is tired that day or in a bad mood that day or whatever, their whole view of that case can change. And then on top of that, hiring these expert reviewers to do this, whether you’re the insurance company or the manufacturer of a clinical trial, is very expensive to do. So what we came away with is that this is a real opportunity to try to improve on a system that will allow for better diagnoses. And so we came up with this concept of InCircle. And InCircle is this based on our study, but what it does is it takes the criteria and puts it into an algorithm that then allows you to make sure that you’re finding the right patient for the right drug. The downstream effects are enormous. Number 1, these drugs are very expensive, so you’re preventing patients from getting expensive drugs that they don’t need. Number 2, these drugs can have side effects. I’ve had patients have strokes from these drugs or renal failure from these drugs. We don’t want to expose patients to risk if there isn’t a potential upside to this. And we think that there’s an opportunity over time to really educate the providers so that they understand the criteria better. So in this space, we’re really the first-to-market technology solution to standardize it. And our approach to this is we want the decision-making to be consistent so you don’t have two world experts coming up with different answers, and we want it to be transparent so that it’s not a black box where doctors send in information and you get some answer and you don’t understand why that answer was derived. So we take the information that the doctors use to make the diagnosis; we organize it, which is probably the most important thing that we do; we can then analyze that data in a number of different ways. And one of the great things about it is we could use the same algorithm for a manufacturer clinical trial, and then also for an insurance company’s prior authorization process, because all we have to do is change the rules. Each company can have a different set of rules and we can apply those rules to the data once it’s into our algorithm. And then we can synthesize that information and then prioritize who should be treated. So the general premise is this, that the records come in. As the records are analyzed, we put people into one of three buckets. So green means, yes, they meet the criteria for this diagnosis, they should be treated. Red means they actually meet the criteria for a different disease that they should not be treated with. And then yellow, which is a common finding here, is we’re not sure. Patients don’t always read the books. Patients don’t always fall within the nice little buckets of red and green. So there are many patients that might be yellow where an expert reviewer could say, yes, let’s try to treat this patient, but we have to do follow-up. And this was a second thing that we really noticed from these insurance reviews, which is once the insurance company did the prior authorization, spending hundreds of thousands of dollars a year to treat these patients, there was no continued analysis of whether the patient was benefiting. So we understand that there may be some patients we treat that don’t benefit and some patients we treat that do benefit. So if they’re not benefiting, we want to get them off of therapy. And that’s this reevaluation process that then allows us to continue to refine each patient’s journey along this path. So again, we want to say, did that patient improve? If they did, then they kind of move, let’s say from a yellow bucket into a green bucket because we’ve now seen that that is the right drug for the right patient. And so this is sort of just a very quick view of how we approach this. So the payers or the doctors can log into our system. They’re asked for a series of required documents. We want to make this as simple as possible for the physician on the end of either enrolling for a clinical trial or trying to get prior authorization by the insurance company. Because then what we do is we have a series of extractors, which are trained nurses that will go through the records, enter that information into our algorithm. In the long run, we’d love to think about an AI system that can actually read the language. So the more doctors move to EMRs, we could eventually start to take out that data extractor piece because it would just go straight into the algorithm. For now, there are too many different systems and we haven’t been able to develop that quite yet, but that’s sort of one of our next steps. The algorithm then analyzes these patients in a number of different variables. These variables, again, are defined based on whatever the payer wants, whatever the manufacturer wants. We can tweak those very easily because we have all of the data already set into our algorithm. And then that algorithm gives us one of these three outcomes. Green tells us, yes, this is a diagnosis, in this case, CIDP. The patient should be treated. Red means there’s something else that we can detect. So for example, we might say that’s diabetic neuropathy and therefore they don’t get treated with immune modulating drugs. And then yellow at this point means we’re not sure. They have some features to suggest that this is the right patient, but they have some features that aren’t consistent. And then at this point, that still goes to an outside reviewer. The goal is over time, as we do this, and as we get more and more patients where we see the outcomes, the number of those cases will get smaller because the system will learn. The system will learn what variables make a yellow likely to respond or a yellow not likely to respond, and we can start to define those better over time. And this is actually one of the very nice things, again, imagine if you’re a payer now and you’ve got hundreds of patients that are receiving these expensive therapies and you want to say to yourself, well, what does my portfolio of patients look like? And so we can give that to the payer or to the manufacturer and say, well, here’s its top case. For example, they were treated for six months as a yellow case. We weren’t sure if they were getting better, but by six months, the notes clearly state that the patient’s better. So that patient moves to a green case. If you look at the third to the bottom, you see a yellow case that’s treated for four months. At the end of four months, it’s reviewed. It’s clear they’re not getting better. So they get a red and they would stop therapy. And these were that start one color and turn to red. In our current system, these patients can be treated for a decade or more. So millions of dollars can be spent on these patients when there’s zero evidence that they’re having any objective improvement because there’s very little oversight of whether or not the patient is in fact benefiting from their therapy. So we did, using our algorithm, just a little test case. So we took 41 cases from a national insurance company, ran them through our system. And the nice thing about these cases was because they had been followed for a while, we could not only say, what was the initial decision? Was the initial decision by the insurance company correct or not? But we could also say, if we had made that case yellow and we had followed that patient over time, when would we have stopped them if they didn’t improve? And so in just those 41 cases, the average spend was $7.8 million — or not the average, the total spend was $7.8 million in the cases that we reviewed. Had they used the InCircle system, there would have been $3 million of savings. So about 40% unnecessary IVIG spend in a market, which I think the last time I looked, at least, was about $6 billion in the US. So this is a massive market. So we’re very hopeful that using these kind of AI systems first in CIDP, but we’re already moving into other areas like immunology, we think we could absolutely move into multiple sclerosis. And we think there’s a very big market to really, again, just have consistent, transparent decisions being made about the treatment of Alzheimer’s disease. Because at least in Arizona, I woke up this morning on LinkedIn, they’re already staging a protest down at the statehouse to make sure that everybody gets access to these drugs. And it’s not that we don’t want to give people the drugs, but for example, in the case of the Alzheimer’s drugs, there are significant risks. Patients have brain edema, small hemorrhages. We don’t want to give it to people that aren’t going to get better. And that all starts with making sure that we know that the patient has the right disease and not just relying on a diagnosis that anyone can make.