Machine Learning for medical tasks with big data

Today the LANIT-TERCOM team faces the task of pathology detection from fluorography. Our developers have been provided with several thousand anonymized digitized fluorographies. Some of them are labeled as fluorographies with pathologies.

The team is working on a solution that identifies abnormalities on fluorographies. This task can be solved using deep learning method. In the first stage, the plan is to filter out deviations from the overall chest pattern. Then analyze the deviations themselves. Based on this, it will be possible to offer a software solution that will significantly facilitate the work of doctors, as it will allow filtering out fluorographies of healthy patients and emphasizing the attention of medical specialists on cases with obvious deviations.

A similar task is also being solved to develop a new method of intelligent analysis of ultrasound monitoring data for automatic diagnosis of cerebrovascular diseases.

The project has an international character and is realized with the participation of the Chinese side, represented by a scientific group led by Professor Minghue Ding.

The methods of intellectual data analysis developed within the framework of the project, in addition to being used in solving the problem of automating diagnostics based on ultrasound monitoring data, can be used in a variety of areas. In case of successful testing, it is planned to realize a hardware platform based on FPGA.