Img.1 An example of service inference
At the moment, the functionality of the service allows you to detect suspected pathologies, make their graphical localization on the analyzed images, and also classify the detected pathological area.
The pathology classifier of the FtizisBioMed service is able to distinguish and catalog 9 types of pathologies (Pleural effusion, Pneumothorax, Atelectasis, Nodule, Infiltration / consolidation, Multiple nodules, Cavity, Calcinate / calcified shadow, Rib fracture).
In addition, the service gives a numerical value for the probability of the presence of pathology. The localization areas are marked with the help of contours, the color schemes of which correspond to the probable classes of pathological signs given in the caption to the image processed by the AI service.
From 2018 to the present, the FtizisBioMed service has passed several independent tests, including the service was tested at the site of the State Budgetary Healthcare Institution "Scientific and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Department of Healthcare" («Moscow Radiology»).
НScientific articles on assessing the accuracy and effectiveness of the proposed service have been published in peer-reviewed journals indexed by Scopus and / or Web of Science:
- С. П. Морозов, А. В. Владзимирский, Н. В. Ледихова, И. А. Соколина, Н. С. Кульберг, В. А. Гомболевский «Оценка диагностической точности системы скрининга туберкулеза легких на основе искусственного интеллекта» Туберкулёз и болезни лёгких, Том 96, № 8, 2018
- V.I.Klassen, A.A.Safin, A.V.Maltsev, N.G.Andrianov, S.P. Morozov, A.V.Vladzymyrskyy, N.V.Ledikhova, I.A Sokolina, N.S.Kulberg, V.A Gombolevsky, E.S.Kuzmina «AI-based screening of pulmonary tuberculosis: diagnostic accuracy» Journal of eHealth. Technology and Application, Volume 16, Number 1, November 2018, ISSN: 1881-4581
To create a modern system for combating the epidemics of tuberculosis and COVID-19, the Russian government organized an Experiment on the use of innovative computer vision technologies for medical image analysis and subsequent applicability in the healthcare system of Moscow.
The FtizisBioMed service participates in the Experiment and, based on the results of 2020 and 2021, is recognized as the leader among other AI services for the processing of fluorograms.
Img.2 Results of the Experiment for November 2021. (https://mosmed.ai/)
According to the results of the official calibration testing of the latest version of the FtizisBioMed service by the Moscow Department of Health (Protocol of calibration testing of the AI service No. 13 dated 24.03.2021), the following results were achieved: area under the characteristic curve (AUC = 0.95), sensitivity 92%, specificity 92 %. Thus, the total number of errors (including false positive and false negative) is no more than 8%.
Since 2019, the FtizisBioMed cloud service, with the support of the Administrations of 53 regions of the Russian Federation, has been used by doctors of X-ray departments of medical institutions of all levels.
Img.3 Map with regions adopting medical AI service
In the process of developing the service, registration documents were obtained that establish intellectual property rights:
1. Свидетельство о государственной регистрации программы для ЭВМ 2017615663 «Программа автоматизированного анализа флюорографических изображений»,
2. Свидетельство о государственной регистрации программы для ЭВМ 2018615227 «Программа для ручной разметки патологий на флюорографических снимках»,
3. Свидетельство о государственной регистрации программы для ЭВМ 2018615226 «Программа автоматизированного анализа флюорографических изображений (версия 2)»,
4. Патент на изобретение 2684181 «Способ автоматизированного анализа ЦИФРОВЫХ флюорографических снимков»,
5. Свидетельство о государственной регистрации базы данных 2020620858 «Размеченные результаты флюорографии».
June 21, 2021 FtizisBioMed LLC became the winner of the National Award in the field of entrepreneurial activity "Golden Mercury", established by the Chamber of Commerce and Industry of the Russian Federation, in the category "Innovative Activity".
Img.4 «Golden Mercury» statuette
On October 6, 2021, FtizisBioMed LLC became the winner of the IT Stars 2021 award named after George Gens in the nomination "Innovative project in Healthcare".
Img.5 «IT stars» award winner statuette
On November 16, 2019, FtizisBioMed LLC filed an application for registration of the "Program for automated analysis of digital chest X-rays/fluorograms" as a medical device.
The FBM service was the first to undergo a clinical trial process using the unique methodology of the NPCC DiT DZM. The certificate of evaluation of the results of clinical trials 001 was received on 27.10.2021.
Due to the change in the legislation on registration of medical devices (MI) from 1.01.2022, the registration of MI was somewhat slowed down due to the large number of applications under the old legislation. Nevertheless, LLC "FtizisBioMed" received an official decision to assign the status of a medical device to the FBM service on 31.05.2022 under the number RZN 2022/17406.
Img.6 Registration certificate for a medical device
The software product automatically evaluates the probability of pathologies on radiographs and a fluorogram of the chest organs. Searches for and localizes pathological signs. It is used in the medical decision support system.
The license rights belong to "FtizisBioMed" LLC
Clinical problem to be solved: Identification of signs and cataloging of pathologies of the chest organs
Instructions for using the FtizisBioMed cloud service are available at: https://ftizisbiomed.ru/.
To access the full functionality of the service, it is necessary to conclude an agreement for the use of the software product with FtizisBioMed LLC. Within the framework of the agreement, the User also has the opportunity to conclude additional agreements with FtizisBioMed LLC regarding training medical personnel to use the service and / or organizing telemedicine consultations with leading specialists of Russian Federal medical centers.
• probability of the presence of pathological signs of chest organs;
• localization of pathological signs of chest organs.
Adult patients over 18 years of age
Area under the ROC-curve (AUC) - 0,96
Sensitivity - 94%
Specificity - 93%
Accuracy - 94%
Analysis duration of 1 study - 1 s
Area under the ROC-curve (AUC) - 0,965
Sensitivity - 92%
Specificity - 94%
Accuracy - 93%
Analysis duration of 1 study - 10 s
Target population and area of use: Adults (over 18 years of age) undergoing routine fluorographic screening, as well as having a referral from a doctor for a chest x-ray for suspected disease, or monitoring the effectiveness of treatment for changes in the lungs. AI-service "FtizisBioMed" specializes in the automatic analysis of images of chest x-rays (fluorograms) in order to identify and localize pathological changes in the lungs.
Purpose of the Software Product: Analysis of digital radiographic images of the chest organs and identification of signs of pathologies of the chest organs.
The service is a support system for making medical decisions.
- Children's age (under 18 years old);
- Non-anonymized data (the presence of the patient's full name);
- Artifacts from foreign objects at the research level, which are superimposed on the lung area;
- The position of the patient's body, which differs from the position regulated for obtaining a direct projection;
- Research performed in any phase of the respiratory cycle, except for holding the breath at the depth of inspiration;
- Study modality, different from radiography / fluorography;
- Anatomical area of study, which differs from the chest organs in full (from the tops of the lungs to the pulmonary sinuses);
- Projection of the study, different from the antero-posterior or postero-anterior;
- Data outside DICOM 3.0 format or images outside of JPEG, PNG, DCM formats.