Diagnostic Robotics data predict spread of virus

Kira Radinsky  / Photo: Eyal Izhar, Globes

Dr. Kira Radinsky has switched her data mining algorithms focus from consumer behavior with eBay to tracking Covid-19.

Nobody listened when Diagnostic Robotics chairman Dr. Kira Radinsky predicted that the new virus discovered in China would paralyze the world, even though she is one of Israel's most famous predictive data mining researchers. But within several weeks they were listening.

Radinsky, still only 33, already has an impressive curriculum vitae. The Ukrainian-born Israeli computer scientist, inventor and entrepreneur co-founded SalesPredict, which developed predicting data mining algorithms to leverage information about online consumer behavior. SalesPredict was sold to eBay in 2016 for $40 million and Radinsky became eBay Israel's chief scientist and director of data science.

However, from 2017 Radinsky began focusing on medical prediction as well as consumer behavior, founding Diagnostic Robotics together with CEO Yonatan Amir and Engineering Director Prof. Moshe Shoham. Radinsky's first prediction was to surprise the medical community by correctly predicting a cholera outbreak in Cuba.

Over the past few months, Radinsky has been focusing her full attention on the coronavirus with Diagnostic Robotics fighting the disease with a triage and monitoring system for healthcare providers and government agencies. The AI system developed by Diagnostic Robotics has been adopted by Israel's Ministry of Health as its basic tool for working with the public, among other things, to map and predict the foci of the virus's spread. Radinsky believes that using this system, the Ministry of Health can decide where to tighten and where to ease the lockdown. The decisions currently being taken regarding this or that town or neighborhood are based on Diagnostic Robotics AI system.

"The proof is in the results and the truth must be said - we didn't expect this virus," Radinsky tells "Globes." "What's the difference between intelligence about previous viruses that we successfully predicted? Those viruses had already appeared a number of times in the past and various places around the world, faded somewhat and we actually predicted their renewed outbreak because the conditions that led to previous outbreaks recurred. In contrast, a virus that has never ever occurred in the past is more difficult to predict. But a month before the virus had reached Israel, we had already decided to adapt our existing systems to track its spread in the country.

Even before Diagnostic Robotics joined the fight against Covid-19, it was already a successful startup with 100 employees in Israel, including 30 physicians and most of the rest engineers, scientists and IT specialists. The company had already picked up customers like governments and hospitals. In Israel, Diagnostic Robotics has worked on pilot projects with a large part of the health organizations such as the Meuhedet and Leumi health funds and hospitals like Soroka in Beersheva and Rambam in Haifa.

Amir says, "The concept behind our system before the coronavirus was more or less this: patients contact the health organization either from home or in the hospital and the system already starts questioning and diagnosing them, and taking down all their personal details, which lets them have access to their personal medical file. When the patient comes to the doctor, the system gives their doctor recommendations - what to examine what to check might have been forgotten to ask, what the diagnosis means in this situation and what prognoses were perhaps missed and how they can be discounted. It's even possible to receive from the system referrals that the patient is likely to ask for and conduct them even before the patient comes to the doctor, thus saving one visit. The doctor can see this information before the visit and even make recommendations via a video call or telephone asynchronously."

Since the coronavirus began, the system has been operating anonymously and its role is to question the public about symptoms in order to understand the spread of the virus in regions and among certain communities. Radinsky says, "The aim is to support decisions such as where to ease and where to tighten the lockdown and where to send ventilators."

Radinsky claims that at the moment in Bnei Brak, their AI system sees not gap between the number of those diagnosed and the numbers reporting problematic symptoms, whereas in Eilat and in the Arab sector there is a gap. In other words such a deviation shows many undiagnosed cases. Diagnostic Robotics symptoms questionnaire system cannot say who is ill but it can indicate a shortfall in normal local morbidity and locate new morbidity in certain neighborhoods and cities.

The company says that it is currently beginning to develop the ability to predict how many cases there will be each week in every city in Israel and over the next four to seven days. Radinsky says, "We use information that is available to the public such as density of population, data about travel between cities, and the distance between cities to identify which city will influence other cities. At the same time, we have the history of morbidity in every region provided to us by the Ministry of Health."

Amir adds, "When the coronavirus outbreak became significant, we understood that we had a unique opportunity because we have built as technology that has all the components that can provide value in this field. We made relatively small adjustments to the system and we presented it to the Ministry of Health, who selected us as its main technology supplier for mapping coronavirus in Israel. We provide the technology to the Ministry of Health, health funds and the Magen David Adom emergency services and it's all for free."

How do you identify in all the data that you receive who is ill with the coronavirus? After all the symptoms are very similar to other diseases, and the virus is so new that we still don't even know its exact clinical picture.

Amir explains, "We are learning things as we go along with the process just as AI does. But we also got a lot of information from China and Italy and other countries, and of course we've scanned all the medical literature that is currently being written."

Whoever fills the questionnaire is given recommendations about what to do?

"No, it is an anonymous, voluntary questionnaire, and whoever fills it in does not get a personal recommendation. They are given comprehensive and official information from the Ministry of Health website."

How many days ahead can you predict?

Radinsky says, "At the moment we can see three or four days ahead. We saw how the ultra-Orthodox leapt in our system just after Purim. My estimation today is that the source of the massive infection in Bnei Brak was due to Purim and not all sorts of improvised prayers after that. Now the number of infections is high because they already have a lot of sick people, and also perhaps because they were a little less careful, but we don't see the spread of infection there, the way it was after Purim."

"The more information that we gather, so the bigger number of days ahead that we will be able to predict. At the moment we only have full information for the past two weeks. In order to predict two weeks ahead, we need at least five times this window - in other words going back ten weeks. The more data that we will have, so our predictions for the near future will improve, and the more data we will have, we will able to offer advice on where to tighten the closure and also where it is possible to begin to open up."

What are you learning about the virus itself from the information?

"We are seeing how the disease is developing. There are patients who develop a fever and then respiratory problems. Others begin immediately with respiratory problems, without fever. Our application confirms the loss of the sense of taste and smell, even up to 30 days from the start of the infection, as well as the gap of 4-5 days between the estimate infection and the appearance of symptoms."

Amir continues, "We see that people who have a cough and fever correctly identify themselves as being sick with the coronavirus. But among those ill with a runny nose and sore throat, there are more people who think that they are healthy but it turns out that they are in fact sick. It's logical because they are without the more common symptoms of the coronavirus but our system has identified that sometimes these symptoms of the disease can appear and be deceiving."

"The aim is subsequently to refer examinations of those who are highly likely to be sick and not known to the Ministry of Health and also to predict a deterioration. This will happen after we have gathered much more information about the disease's development."

What does your system say about when an if we will have a shortage of ventilators

Radinsky stresses, "At the moment we don't foresee a shortage of ventilators but there is a shortage of manpower to operate them."

When will Covid-19 end?

Radinsky responds, "I'm unable to answer that question. The virus is new. One model is that of SARS, which was over after six months, perhaps because it came to an end. We don't know but we see that in Wuhan in China, it took about four months to gain control over the rate of infection. Probably because among other things herd immunity was created and so that's about the scale of that."

In contrast the Spanish flu persisted for two years in two waves, regardless of the weather. Some people caught that virus several times after it underwent a mutation. If after six months we see no easing up as a result of the complete lockdown, then probably we're talking about the second model. The fact that there aren't so far many new mutations and immunity seems to be generated after the disease means perhaps we are more similar to the first model. But the policy of having most of the public in lockdown and not being infected can lead to an additional wave when the closure is lifted because immunity hasn't been created. In short, we don't have an answer yet."

Is it possible to use your data to see which exit strategy will be best?

"Evaluating an exit strategy is not our work. It is a very interesting problem but it is being handled by the Ministry of Health, which can rely on, among other things, the data that we are providing them."

Your system can be relevant for monitoring the population and early detection of outbreaks in the more distant future, after the closure is lifted, on condition that Israelis continue to report on a day to day basis. Is there a chance that this will really happen after the clear and immediate danger has passed?

"I think there will be a psychological change. People will move away from the lives they have been leading until now. We feel that some of these habits have already been assimilated, for example reporting various symptoms that you have and to frequently and continuously monitor your medical condition, and it is worthwhile also closely monitoring the public health situation. Perhaps that will be the new world in which a new medicine will emerge."

Radinsky continues, "Viruses expose our biggest problems as a society. Just as cholera exposed the need for clean water, so the coronavirus has exposed the powerlessness of the health system to provide the best personal treatment for everyone who is sick. There will be no choice and it will be solved by digital healthcare."

In answer to the question if the company plans making their data accessible to all researchers and health systems to speed up research into Covd-19, Amir said that the information has been transferred to the Ministry of Health and they are responsible for making data available to researchers and the public at large, according to their considerations. He said, "The aim is to make all the data accessible to the academic community so that everybody can study the virus in the best possible way."

Published by Globes, Israel business news - www.globes-online.com - on April 10, 2020

© Copyright of Globes Publisher Itonut (1983) Ltd. 2020

Kira Radinsky  / Photo: Eyal Izhar, Globes
Kira Radinsky / Photo: Eyal Izhar, Globes
Twitter Facebook Linkedin RSS Newsletters גלובס Israel Business Conference 2018