Mathematicians and Practicing Surgeons to Fight Venous Diseases

© iStock

One million people in Russia suffer from venous diseases. The ‘Intelligent data analysis for healthcare information systems’ Mirror Lab project brings together expertise in mathematics and medicine in order to better diagnose various conditions in phlebology. Project leader Vasilii Gromov talked to The HSE LooK about its achievements and prospects.

Vasilii Gromov, Professor at the School of Data Analysis and Artificial Intelligence, Senior Research Fellow at the International Laboratory for Intelligent Systems and Structural Analysis, and leader of the project ‘Intelligent data analysis for healthcare information systems’

Why Doctors Need Mathematicians

Ideally, any mirror lab project would be interdisciplinary and combine two fundamentally different groups of researchers. In our case, these groups are mathematicians specialising in artificial intelligence and practicing surgeons who have little interest in our mathematical models but need a solution for their specific problem. There are tasks where applied mathematicians are not needed because a method already exists and just needs to be applied. However, there are also tasks where our presence is necessary, ie when the existing methods will not work. Diagnosis of venous diseases is an example of this type of task.

Our common challenge was learning to speak each other’s language

Our colleagues tried to understand why some of our formalities are important. For us, mathematics is a way of describing reality; if something doesn’t work, then this means that the chosen method is not applicable to the given task. We tried to understand their requests and their perspective of what they consider right and wrong. This kind of interdisciplinary experience is important, even if it slows down the process because too many things must be understood overall.

Mirror Labs are joint research projects between HSE University and its partner institutions. Launched in 2020, this format was the first inter-university initiative of its kind aimed at developing networks and collaborative partnerships, while also enhancing cooperation in research and academic activities between HSE University and other leading Russian academic institutions.

The name ‘mirror’ labs refers to the financing of projects on parity terms and pooling research resources from both sides: providing each other with necessary expertise, sharing analytical methodology and results, conducting training seminars for each other and hosting interns. Research activities may be carried out by joint project teams over a period of three years with the option for extension.

Mirror Lab projects can be carried out in various fields and a diverse array of formats.

Advances in Diagnostics with Machine Learning

While conversing with various colleagues (physicians, phlebologists, doctors of science in this particular field, etc), I heard a terrible number:

One million people in Russia suffer from venous diseases

The situation is dire and, as such, the task was to understand how the disease proceeds for particular people, ie, which vein gets blocked, which vein doesn’t get blocked, which vein gets partially blocked, and whether surgery is necessary or not. All of these factors can be learned from certain parameters and characteristics of the system, such as cross-sectional MRI, venous flow velocity in superficial veins, computed tomography, and so forth. From a mathematical point of view, the situation looked quite simple: we had to solve the so-called inverse problem of mathematical physics with the application of neural network approximation. However, it turned out that the body of healthcare data is actually very small.

While classical machine learning problems involve dealing with millions of pieces of data, here, we have around a hundred

We identified three interrelated approaches for solving the data problem: data collection from colleagues who deal with phlebology issues, sample augmentation using special mathematical methods, and the application of physics-informed neural networks.

The second task is no less important from a social and medical point of view. This concerns fistulas, artificial tube-like connections used particularly for patients who go through hemodialysis. The problem is that fistulas tend to get clogged and can lead to blood clots breaking off and restricted blood flow. With this in mind, we developed a device that can be attached to a fistula with a patch and identifies different types of blood movement inside it: ‘normal’, ‘consult a doctor immediately’, and ‘not critical yet, but you should probably consult a doctor’.

An Inter-Regional Project Becomes International

A former member of our department used to be Vice Rector for Science and Innovation at Surgut State University. When he enquired about which innovations could be introduced at his university, he learned there was a team of rather strong doctors and started looking for matching specialists to work with. So, we met, talked, and found some common points of understanding.

Today, our project team consists of specialists in mathematical modelling and AI from HSE University, and physicians, surgeons, diagnosticians, and specialists in social medicine from Surgut State University.

Moreover, our project attracted various specialists who deal with the same issues. As such, doctors from St Petersburg, Nizhnevartovsk, Samara, Tashkent, and the Netherlands are actively cooperating with us, and we hope that this list of people—to whom I would like to personally express my gratitude—will expand, and we will become a point of attraction in a broader sense.

We are progressing in phlebology, but the implementation period is much longer than three years

There have been practical developments, and we definitely know how to build a mathematical model. However, we need more human resources to complete it. Right now, we lack software engineers, mathematicians, and people who can organise data collection, but of course, we will continue working on the project!