ITMO: Machine learning technology for diagnosis of inflammatory diseases of intestines

Main articles:

• Machine learning (Machine Learning)
• Artificial intelligence in medicine
• Big Data in medicine

2020: Victory of the Russian algorithm in international contest

On September 8, 2020 the Philip Morris International company announced Zdrav.Expert that young Russian scientists won international contest on creation of machine learning technology for diagnosis of inflammatory diseases of intestines.

The Russian algorithm of diagnosis of inflammatory diseases of intestines the Source is recognized the best on international contest: intervals.science

Artem Ivanov and Vladimir Ulyantsev from Computer Technologies laboratory of the ITMO University created an algorithm which on the basis of noninvasive methods of diagnostics with the high level of reliability reveals presence at the patient of inflammatory diseases.

The Russian team became only, won at once two stages of tender. At the first stage of tender scientists of ITMO became winners along with commands from the Medical university of Georgetown (USA) and Institute of the high-performance systems of Italy. At the second stage the Russian team won a victory together with representatives of the University of Luxembourg and the Paduan university (Italy).

Competitions MEDIC. Source: intervals.science

Tender, with the general prize fund of 12,000 dollars, took place on the crowdsourcing sbvIMPROVER platform created by research division of Philip Morris International. In total was 81 applications from 15 scientific commands from the different countries of the world are submitted.

In Computer Technologies laboratory of the ITMO University we develop more than seven years algorithms for the analysis of metagenomic data. Generally we apply them to data of a microbiota of intestines: we carry out contrastive analysis, we analyze antibiotikorezistentny genes. Certainly, it was interesting to us to look how our algorithms will cope with a problem of diagnosing of inflammatory diseases of intestines, - Vladimir and Artem told.

As the competitive task having independent scientific value search of an optimal algorithm of machine learning was defined unmistakably to diagnose differences from each other of ulcer colitis, a disease Krone, inflammations of a rectum and other diseases. As initial data scientific commands received results of noninvasive methods of clinical diagnostics and had to apply analysis algorithms of "Big Data" to computer processing of results.

At the first stage provided to participants data of metagenomic sequencing. At the second stage of a command received previously calculated matrixes of taxonomical data and the metabolic ways given to recurrence. It allowed specialists to compare without access to methods of the analysis of metagenomic data efficiency of classification beyond the scope of preprocessing.

The solution found by a team of the Russian scientists will allow to classify with the high level of reliability inflammatory diseases of intestines at patients.

We could reveal strong and weaknesses of our algorithms and to compare their accuracy to approaches of other participants. But with a victory in tender work is not finished, and we will continue researches for more complete description of signs and more exact diagnosing of diseases, - Artem Ivanov told.

it is very important For us that a data set for machine learning and test data were taken from two different researches that it was possible to be convinced that the received result is not inherent to only one research. Recently we study a microbiome as it is connected with many diseases and in general is responsible for the state of health. For us it was important to be convinced that the methods developed during tests precisely define structure of a microbiome. – Stephanie Bu, the manager on ensuring scientific transparency and verification of data of PMI told. According to her, further the project will cover other areas of biomedical researches.