Scientists from AltSU participate in the development of classifiers for medical images of CT lungs screening based on artificial intelligence

2 April 2022 Department of Information and Media Communications
The team of the Department of Informatics of Altai State University, together with representatives of Altai State Medical University and the Diagnostic Center of Altai Krai, are working on the development of medical image classifiers based on artificial intelligence. This is part of a large project "Development of technologies and software products for the health of the population", which is implemented as part of the "Priority 2030” program.

The project is led by Denis Kozlov, Head of the Department of Informatics at the Institute of Mathematics and Information Technologies of AltSU.

The classifiers will be used in the interpretation of CT scans of lungs. In particular, in the analysis of spherical formations in lungs, which, as a rule, have three causes: cancer, pneumonia or tuberculosis. To date, it is possible to reliably identify the origin of the formation only with the help of histological examination, when a tissue fragment is examined under a microscope. This study can bring discomfort to the patient, and also takes time. And in some cases - for example, with oncology - it is impossible to delay treatment. This is where artificial intelligence comes to the rescue.
The task of scientists is to create software that can analyze and classify the obtained CT images.

They are currently building models using a typical programming task, the Supervised Machine Learning Problem. The computer is "shown" a lot of images (in this case, lung formations), and then the correct answers - in which case, what was diagnosed according to the results of the histological examination. The task is to teach the neural network to identify patterns. And when it is given a new image that has not been shown before, it will be able to identify, based on the accumulated experience, which class is represented on it: tuberculosis, cancer or pneumonia.

Similar work is being done by scientists around the world. This has been made possible by the enormous advances in artificial intelligence that have been made in the last decade. But here scientists are faced with the problem of anonymity of personal data. The Law on the Protection of Personal Data does not allow using them even for research purposes.

“The team of the Department of Informatics, together with colleagues from ASMU and the Diagnostic Center of Altai Krai, had to solve the problem of data depersonalization. All information about the patient when performing a tomographic examination is stored in a file of a special DICOM standard, and, as a rule, stores a number of personal data. Before passing them on to a researcher, they must be depersonalized. We wrote special software that can work with files of this format and delete this personal data. The images themselves remain intact, as does the information about the diagnosis, and we can use them to build such a classifier that would be able to distinguish between classes of diseases on a CT image,” says project leader Denis Kozlov.

The implementation of the project will improve the efficiency of the university's research processes through the introduction of artificial intelligence technologies and big data processing methods into research processes, the development and implementation of technologies and products to ensure the quality of public health and the prevention of socially significant diseases. In addition, the development of the infrastructure for the development and promotion of high-tech tools for diagnosing and predicting the clinical course of oncological diseases aims to ensure the transition to personalized medicine and the development of new health-saving technologies.

Printable version