Page 153 - AIH-2-3

Artificial Intelligence in Health Explainable solutions from AI for HSSs

Figure 4. The model of the KESs manufacturing environment
Abbreviations: API: Application programming interface; IACPaaS: Intelligent Adaptive Clinical Platform as a Service; URL: Uniform resource locator.

to compare it with the result, fixed in this precedent. It tools are added to the development environment (as a
is advisable to develop a tool for checking the quality of framework) one after another.
the knowledge base together with a solver since they have We compared the process of building a complex of
many common software blocks. interconnected evolving knowledge bases to make medical
If it is necessary to update the knowledge base in systems using the Protégé and IACPaaS tools. We created
6
30
connection with obtaining precedents that do not classes of diseases, symptoms, and drugs on the web, in
correspond to the knowledge (if the precedent contains Protégé. Next, we had to associate diseases with symptoms
the correct result of solving the problem, which does not using the object property mechanism (some acute diseases
correspond to the result obtained with the help of KES), required a dynamic description of the clinical picture).
then it is effective to form a new version of the knowledge However, these mechanisms did not describe knowledge
base automatically, based on the methods of inductive in a way that doctors would need and understand, and
formation of fragments of the knowledge base. An explaining the dynamics of disease development was
29
example of such an update of knowledge “from practice” particularly difficult. It should be noted that the Protege
occurs when, after a certain period, the correct result of tools are incomprehensible and difficult for doctors; they
diagnosis or treatment becomes available from a medical are intended for knowledge engineers (although describing
institution. This outcome is then compared with the result the dynamics proved difficult for them). Similar difficulties
from KES: are they contradictory? In such cases, evaluating were encountered when trying to describe treatment
the consistency of the updated KES’s work result with protocols, taking into account the specifics of drug use and
existing precedents is preferable. patient characteristics. Protégé’s mechanisms did not allow
To implement the process of monotonous improvement the patient’s history to be formed as a single document.
of knowledge bases, the following means are required: In contrast, the advantages of IACPaaS in addressing this
(i) inductive knowledge creation tools for each intellectual limitation have been demonstrated.
problem solved (diagnosis, planning, forecasting, etc.), 5.4. Implementing a clinical knowledge-based
(ii) tools to support the choice of precedents (correctly decision support system
solved problems in a statement), and (iii) tools for
verifying the correctness (quality) of the new version The formation of a Medical Portal, Med-IACPaaS (https://
of the knowledge base (the same as described above). iacpaas.dvo.ru/), began with the development of medical
The “knowledge updatability” property depends on the ontologies and editing tools. Previously formed by experts
availability and performance of the above tools; other and knowledge engineers, the medical ontology has been

Volume 2 Issue 3 (2025) 147 doi: 10.36922/aih.5736