The service was developed with the use of automated recognition, machine translation and medical texts analysis tools.

Statistical and registry data processing was made with the use of network-meta-analysis.

  • Libraries and databases of the text data analysis
    Apache PDFBox, Tabula;

    cTAKES , RxNorm_index, OrangeBook, UMLS, LVG;

  • Network-meta-analysis
    Calculations are performed in R programming language (

    The network meta-analysis performs both direct and indirect paired comparisons of the interventions of interest. At the same time, it is possible to select a reference intervention and a subgroup of patients by a risk factor (e.g. decline in kidney’s estimated glomerular filtration rate).

    The criteria for including a study in the network-meta-analysis:
    • randomised controlled trials (RCTs) and verified registry data that are based on minimum 2 groups. The comparison group can be either a placebo or an active control;
    • availability of clinical outcome data for calculating the measure of the safety and effectiveness of an intervention (effect size);

    Depending on the data format the incidence rate ratio (IRR) or odds ratio (OR) is used as a measure of the effectiveness/safety of the intervention, with the corresponding 95% confidence interval (CI).
  • Technology stack
    Programming languages: Java, Kotlin, Javascript, R.

    "Spring" framework is used as a server of applications. Data storage and data processing are done with "DBMS Postgresql".

    All basic components of the software are developed with the help of MSA. Interoperability is provided through message exchange which is done according to the protocol "AMQP" by using "RabbitMQ". The components are put into Docker containers and are managed by "Swarm".

    Interface: React.js + Material-UI.

    Data exchange: GraphQL.

    Authentication and authorization: JSON Web Token(JWT).
  • Hospital information system (HIS) integration (JSON, XML)
    Quick access from HIS to at the step of selecting a patient’s disease.

    Data export from the final step for every patient’s profile saved in A health facility should provide a possibility to import the data and to link it to the patient’s profile in HIS.

    To save data automatically in HIS database drug and disease identifiers are reconciled.

Interpretability of machine learning(ML) outcomes is crutial for scientific community, for developers and for clinicians. Without it a full application of AI in evidence-based medicine is impossible.

MedicBK provides a possibility to check the information accuracy:

- decision making tree – by CPG (ESC & AHA, NCCN)

- medical statistics – links to articles that were used to train ML mode

If you have any questions or you would like to join MedicBK, please fill the form below
Made on