Code prediction: New AI predicts outbreaks of dangerous diseases
MTUCI has developed an artificial intelligence model capable of predicting outbreaks of infectious diseases. Initially, the system was created to monitor the spread of malaria in Angola, but its architecture allows the solution to be adapted for other countries, including Russia. According to the authors of the project, in the future, the technology can be used to predict the spread of seasonal influenza, tuberculosis and other infections, as well as new viral diseases with pandemic potential, similar to coronavirus infection. For more information about how the system works and which diseases can pose the greatest threat, see the Izvestia article.
AI will help predict disease outbreaks
One of the authors of the project came to Russia from Angola to use AI to develop solutions for remote development of the healthcare system in his country. In the course of their work, the researchers found out that the model they created is universal and allows you to customize forecasts for infectious diseases. It is a hybrid AI system that combines machine learning methods, time series analysis, climate factors, and epidemiological data.
— The main feature of the model is the combination of several analytical approaches. This combination makes it possible to identify complex patterns that traditional models are often unable to detect. At the same time, our goal is not just to predict the number of cases, but to provide health authorities with a decision—making support tool for early detection of outbreak risk, optimizing resource allocation and improving the effectiveness of preventive measures," Joaquim Timoteo, a student at the Faculty of Cybersecurity and Information Security at MTUCI, told Izvestia.
During the tests, the model was trained on data on the incidence of malaria in Angola for the period from 2000 to 2025. After that, the system learned how to predict the development of the epidemiological situation for 6-8 weeks ahead.
"With high—quality data, it can be adapted to predict other infectious diseases, including those that are under WHO surveillance, such as coronavirus infections, hantaviruses, dengue fever, cholera, and others," he said.
Respiratory infections, primarily caused by RNA viruses, have the greatest pandemic potential today, as they are characterized by a high mutation rate. These include influenza viruses and coronaviruses. They are able to form new strains against which the population has no immunity, said Andrey Pozdnyakov, an infectious disease specialist at Invitro.
— At the same time, infections transmitted by other routes, as a rule, cause local or regional outbreaks, not pandemics. Antibiotic—resistant bacteria, which can gradually spread in medical institutions, also pose a separate threat, and climate change is contributing to the expansion of the range of carriers of infections such as dengue fever and West Nile virus, he added.
Early warning system
After the COVID-19 pandemic, interest in predictive healthcare systems has grown significantly. Such solutions make it possible not only to respond promptly to the already begun increase in morbidity, but also to assess epidemiological risks in advance, which helps to more efficiently allocate medical personnel, medicines, diagnostic facilities and other resources of the healthcare system, said Artem Karpov, Chief Operating Officer of Architech II.
— Now the development looks like a promising research prototype. The strong point is the universal architecture, which can potentially be adapted to various diseases and regions. At the same time, for practical application in Russia, the model needs to be trained on domestic data, tested for several epidemiological seasons and tested together with epidemiologists," he explained.
Such systems are in demand in almost all areas where the analysis of large amounts of data is required, said Petr Kshnyakin, deputy head of the laboratory of personal medical Assistants at the NTI Center based at SamSMU. According to him, artificial intelligence is able to effectively identify the risks of outbreaks of infectious diseases, but the accuracy of forecasts directly depends on the quality and completeness of the initial epidemiological data. The more information about morbidity, outbreaks, and the dynamics of their spread can be collected over different periods, the more accurate the model works.
— The architecture of the model opens up opportunities for scaling and adaptation to various healthcare systems. In the future, this technology may find application in Russia to predict the spread of socially significant and seasonal diseases, improve analytical tools and support decision—making in the healthcare sector," said Sergey Erokhin, Rector of MTUCI.
Predictive models in epidemiology work more accurately where there are long series of observations with pronounced cyclicity, said Marina Chumakova, a leading market expert at NTI Helsnet. According to her, analyzing seasonal flu or tuberculosis using AI is easier than HIV. In the case of the latter, forecasting is useful for assessing the burden on the healthcare system and planning the supply of therapy. However, it is impossible to predict HIV outbreaks in the usual sense, since the spread of infection strongly depends on behavioral factors that are difficult to collect in the form of data for model training.
According to Pyotr Kshnyakin, the main limitation of such systems is not the algorithms, but the quality and completeness of the medical data on which they are trained.
Переведено сервисом «Яндекс Переводчик»
