Blockchain and AI Via doc.ai Could Lead to Medical Breakthroughsbr>
Artificial intelligence and the blockchain are combining to advance medicine through doc.ai, a technology platform to generate insights from combined medical data. Built by a team from Stanford and Cambridge University, doc.ai uses blockchain to timestamp datasets and decentralize artificial intelligence. That allows patients to have “conversations” about their health.
As envisioned, the patient can ask the AI questions such as, “What should be my optimal Ferritin value based on my iron storage deficiency?” Or, “How can I decrease my cholesterol in the next 3 weeks?” Or, “Why was my glucose level over 100 and a week later it is at 93?” AI provides the answers and additional context for each answer. Patients can then leverage those conversations with the AI-powered robo-doctor in consultation with their personal physician.
Over the next year, doc.ai expects to roll out three natural language processing modules to medical providers and payors: Robo-Genomics, Robo-Hematology, and Robo-Anatomics.
“We are making it possible for lab tests to converse directly with patients by leveraging advanced artificial intelligence, medical data forensics, and the decentralized blockchain,” said founder/CEO Walter De Brouwer. “We envision extensive possibilities for the use of this technology by doctors, patients, and medical institutions.” An initial coin offering (ICO) for Neruon (NRN) tokens, based on the ERC-20 standard, will commence on Sept. 7.
De Brouwer talked with BlockTribune about the new service.
BLOCK TRIBUNE: Is it accurate to refer to this service as “Robo-doctors?”
WALTER DE BROUWER: Well, I think it is. Although if you compare it to a real doctor, we are very far away. But there are 150 specialties in medicine in the United States. So the more vertical, the more specialized the information, the better the machine gets at it. So we in the company. Rob0-health is certainly on the agenda for the next year or two, because everything that was prophecy five years ago is now happening. We call personalized medicine and precision medicine quantified biology.
All it is is getting the machine to handle these big tables of numbers. You know, it’s a bit like what we did in banking First we had the ATM machines. I think the same will happen in the healthcare. This will be used by medical professionals.
BLOCK TRIBUNE: Walk us through how it works.
WALTER DE BROUWER: We have several use cases in our brochure and one of the use cases is an organization in Burlingame, California. They consist 120 parents. and these parents have one thing in common – they all have children with epilepsy and they come together monthly and they compare results. But most of their results are like an agenda. They have some theories, but they’re not data scientists. They have data, but it’s not uniform, and so they keep on doing this because basically their kids have between 70 and 80 seizures per day.
So one of the parents sends a public a public address to the neural network and gets a private address back, and it says, “I’m honored to be here. Train me.” And this is conversational. It says things like, “Why don’t you let me in with your Facebook account,” and then it responds, “I already know 708 things about her that are very interesting now. Can you also take a picture of her so that I can have more attributes of her face?” It’s actually very magical. It gets your age, your gender, but also your height, your weight, your BMI, and you could correct it, or not. And then it says “OK. Any way you can you go to her medicine cabinet and put her medicine on a table and then take a picture of it?” And so you take a picture and it’s like, “Oh, I see. So she’s on these medications. Where do you go to the pharmacy? Walgreens. OK, I’m at a Walgreens portal now. So if you type in your password (it’s all encrypted), I’ll download all the information.”
BLOCK TRIBUNE: Let’s cut to the chase. What will be the outcome of this?
WALTER DE BROUWER: The outcome is that they have a very structured medical record, which we call quantified biology, of their child. And from that – although they have baselines – the machine can do can set up a model. The model can predict. The model can look for triggers. They can actually compare it to other children who have the same disease. And they can broadcast it for data scientists to launch a mobile on it and the machines will choose the best model. Jeremy Howard is my partner in the center. So they have, basically, a prediction model based on their classified and unstructured data, which is a lot more than they had before. And then they can go to the doctor and say, “What do you think of this? Here is what the machine says.”
BLOCK TRIBUNE: Do you anticipate that there will be new discoveries that are gleaned from all this information?
WALTER DE BROUWER: I’m sure of it, because we tried it out internally already. We came up with very interesting information which we didn’t know about ourselves, things you sort of know, your suspicions about the origin, and other things you would never know. And if you tell a doctor, then he says, “Yeah. Of course.” There is so much information now in the medical sector, doctors have information coming out of their ears.”
BLOCK TRIBUNE: So this is, in a sense, a synthesis of knowledge that will come out with some sort of recommendation about treatment?
WALTER DE BROUWER: Yeah, it will, basically. Also, it will classify the data. It will say, “During the last year, you always had trouble with your triglycerides. We should monitor that. At the same time, I see in your genomic test that you have an IGF deficiency. So that should actually impact your blood result. Let’s talk about this with your doctor.” It might be that he has the wrong medication because of pharmacogenomics. You know, there might be a problem.
BLOCK TRIBUNE: Obviously, you’re going to have to have access to a computer and a lot of hashing power to get this done. How will this work in areas where there is little to erratic electricity? Can that just be done with an app?
WALTER DE BROUWER: Yes. They don’t need the computer for that because we decentralize it. So a lot will actually happen on the phone itself. Also the datasets will be kept on the phone, which is still more difficult, but it’s not impossible. But training the neural nets that for the moment is not possible. For that, we will have to go back to the cloud.
BLOCK TRIBUNE: You are planning an ICO.
WALTER DE BROUWER: So the tokens we call Neurons, because basically, it’s the product of machine learning. It’s basically prediction. There are 86 billion neurons in our head.
BLOCK TRIBUNE: So all the information will be based on a blockchain. Is that correct?
WALTER DE BROUWER: Yes. Because that’s what allows us to decentralize, to push it to the edge. Also, time-stamped data to use cryptography and to tokenize everything, how far out in the future are we looking for this to be a regular tool used by medical professionals. It is basically ready so we can open the network in less than a year. But then there is the adoption, and there I’m talking to several corporations and patient advocacy advocacy groups. Because once you arrive at 100,000 users, then all the rest takes care of itself you know. But you have to have the first 100,000, I say.