SYN

A predictive analytics tool for strokes

Alviss.ai CEO and Founder Dr. Pavel Overtchouk talks about how the company’s stroke prediction software seeks to be the first attempt to apply predictive medicine to cardiovascular interventions and heart surgery.

Alviss.ai CEO and Founder Dr. Pavel Overtchouk talks about how the company’s stroke prediction software seeks to be the first attempt to apply predictive medicine to cardiovascular interventions and heart surgery in response to emailed questions.

Why did you start this company?

Heart surgery replacing the aortic valve is a lifesaving intervention that extends life expectancy by 15 years or more in elderly patients presenting with stenotic aortic valve calcification. About 5% of the population over the age of 65 present this condition. Transcatheter replacement of the valve works in most cases and patients get a new functioning bioprosthetic valve reducing their shortness of breath and chest pain. But a significant proportion of patients will unfortunately develop complications during the procedure. Stroke is one of the most dreaded complications. We have observed patients presenting this complication in our interventional cardiology practice. 

Predicting which patients are going to have a stroke has been considered to be impossible due to a complex entanglement of various factors. But we thought that with the recent progress in artificial intelligence, such relationships could be captured and identifying patients at high risk of stroke could become reality.

Hence I developed an AI-powered software using patient anatomy and procedure data to predict which patients are at high risk of stroke. Encouraged by our scientific results, I decided to build a commercial product out of this software to make it accessible for everybody.

What need are you seeking to address in healthcare?

Heart surgery-related stroke causes a major burden for patient health and cost for hospitals. Roughly 100,000 aortic valve replacement surgeries are performed every year in the United States alone, and 3x as many worldwide. Stroke occurs in about 3% of patients. It can cause paralysis, dementia, increase the risk of dying in the days after the surgical procedure. Stroke also is responsible for a high-cost burden for hospitals and health providers, with every stroke costing about $100,000. 

Pavel Overtchouk

A few companies have developed cerebral protection devices. Those are positioned during the surgical procedure in or above the aorta to capture valve debris that embolize during valve replacement and cause stroke. Those devices have vascular complications and extend the duration of the procedure. Using those cerebral protection devices also costs too much to be used in all patients undergoing valve replacement surgery.

Our stroke prevention software would fit in this workflow by identifying the patients at high risk of stroke. Those patients would have the highest benefit in using cerebral protection devices. Hence our product would allow an individual patient-specific approach and act as a therapeutic decision support tool.

Based on our simulations, we estimate that taking this approach using our software to select patients in whom cerebral protection devices should be used, we can reduce by two-thirds the risk of stroke. We could also reduce stroke-related costs for hospitals and public and private payers by about $500 million.

What does your product do? How does it work?

Interventional cardiologists and heart surgeons or their assistants connect to our web application. Then they enter patient and surgical procedure data into our software. Patient anatomy notably about the severity and extension of the calcification of the aortic valve is associated with the risk of stroke. Procedural characteristics such as the transcatheter vascular access and repeated manipulations of the valve during replacement tend to increase the risk of stroke as well. But it’s the overall entanglement of patient and procedural factors that yield an increase in the risk of stroke. In those patients, using cerebral protection devices would mitigate the risk of the stroke occurring. 

Upon calculation the users get a risk estimation, an interpretation of a patient’s risk and guidance for the procedure. Physicians and surgeons remain the guarantors that the results are in accordance with their own experience. But they now have an objective risk assessment they can use to decide upon using cerebral protection devices. We plan on progressively automating the data input process to streamline the risk assessment for users. To do it, we will integrate our product inside of third-party software already present in hospitals.

More than 10 hospitals in Europe and the United States are conducting free trials of our product. FDA approval is pending but we are already looking for partners in hospitals performing transcatheter aortic valve replacements all over the country, especially in Texas, to better understand expectations of cardiologists, surgeons, and hospital decision makers for our product.

Our product, the CerebroVascular Event Risk Calculator (CVERC), is the first attempt to apply predictive medicine to cardiovascular interventions and heart surgery.

Is this your first healthcare startup? What’s your background in healthcare?

No, I have been involved in start-ups since 2018. A previous startup I joined tried to securely bridge patient clinical and pharmacological data to generate insights about medication compliance. That startup failed but I used this experience to build several products of my own making while estimating the product-market fit. 

Predictive medicine applied to cardiovascular interventions is my favorite topic of research and development. I have published many scientific articles on the topic during my medical and interventional cardiology training in France, Switzerland, and Canada. I have treated thousands of patients over a span of 12 years of bedside practice, and got various medical diplomas including in interventional cardiology, echocardiography and intensive care. I have been building machine learning systems and software for 5 years now. Those skills I learned by myself following online courses and resources during my spare time. 

Out of all the software that I have built, the presented stroke prevention software is my first attempt to bring a software-as-a-medical device type of product to the healthcare market. In many ways the product represents the entanglement of the different skills and knowledge I acquired.

What is your company’s business model?

Software-as-a-service (SaaS) is the model we adopted to bring the product to hospitals. 

Interventional cardiologists, heart surgeons or their assistants use either our:

  • Web application: to enter patient and procedure data to get risk estimates, or
  • Third-party software integration: inside electronic healthcare records and computed tomography processing software

The amount we charge hospitals or health providers depends on the number of patients treated yearly or monthly, and is based on their preference.

Who is your customer? 

Under alviss.ai’s B2B SaaS model, our primary customers are hospitals and managed Medicaid health plans.

We also serve individual physicians using the software to estimate the surgical risk of stroke.

We plan on serving on both sides of the Atlantic in the United States and Europe.

How do you generate revenue?  

Access to our software is provided through a paid subscription offered to hospitals. Hospitals are accustomed to SaaS-based software relationships that they have with many providers. Both our web application and third-party integrations will have subscription-based access offers.

Do you have clinical validation for your product?

Yes. The predictive performance of our stroke prediction software has been demonstrated in a published peer-reviewed clinical study, with another study underway. The published work showed in 2021 that CVERC was able to predict four out five strokes with good performance. This international collaboration between France, Switzerland and Japan led by Prof. Thomas Pilgrim, Dr Taishi Okuno and Dr Pavel Overtchouk yielded encouraging results for our product using anatomy and procedural data from about 1500 patients. The scientific article can be freely accessible using this link.

Further clinical investigation is ongoing to estimate the impact of our stroke prediction software for cerebral protection decision making.

At what stage of development is your lead product?

Our AI stroke prediction software is available for free trials using our web application and partner third-party integrations. Readers of this article are encouraged to contact us at pavel@alviss.ai if interested. FDA and CE mark review is underway for commercial deployment in the upcoming months.

Photo: tonefotografia, Getty Images

Shares1
Shares1