Artificial intelligence, deep learning techniques and neural networks based tools in personalization of clinical decision. The original “case-like-this” approach.

The initial hypotheses are usually based on the individual investigator’s pathophysiological knowledge, experience and intuition, meaning that their selection is more casual than it might be if performed systematically, with adequately designed technological tools.

Moreover, these hypotheses are mostly based on single-mechanism interventions decreasing the likelihood of their usability due to biological multi- system relations and biocomplexity. These two factors are among the most significant reasons for the relatively poor efficiency of scientific translation in medicine from bench (preclinical studies) to bedside (clinical trials) when assessed by numbers of proposed biomechanisms or therapeutic ideas that were further confirmed in clinical trials, and were eventually adopted into the clinical management protocols.

Although the Evidence-Based Medicine approach enormously improved healthcare efficiency in comparison with previous intuition and pathophysiology – based management, its protocols still suffer from a few major problems that limit further care development and these points include:

Testing mainly single-mechanism hypotheses or single-therapy ideas, whereas in biology, the

reactions are often secondary to complex multi-system stimulation–inhibition mechanisms.

Ignoring mechanisms that are not straightforward in pathophysiological reasoning.

Testing hypotheses that are selected unsystematically, based on investigator-initiated mode, and thus they are investigator’s individual characteristics depended.

In our project, we address these issues by:

Providing a tool for the selection of unidentified risk factors for the non-communicable disorders of the largest populational burden in addition to those already identified with a standard hypothesis–driven protocols, applying the hypothesis-free approach with deep analysis of individual big multiple origin data in patients with known long-term clinical courses.