OBJECTIVE

To standardize the requirements for describing creation and validation of machine learning models ensuring transparency and confidence in healthcare and regulatory authorities.

MATERIAL AND METHODS

Analytical study of data was based on the Cross-Industry Standard Process for Data Mining methodology (the first four phases of the research cycle were implemented; evaluation and implementation phases should be performed as external independent validation). Analytical methods of scientific knowledge including analysis, synthesis and induction were used.

RESULTS

Practical and methodical guidelines on description of development and validation of machine learning models have been developed. These guidelines are made in the form of methodical document and checklist for accelerated verification of structure, completeness and quality of content. The principal differences between the author’s approach and analogues are standard structure of medical scientific article; balanced presentation of clinical and technological information; inclusion of specific aspects of machine learning that are most relevant from a medical point of view; methodical information to support the authors; ensuring the possibility of applying the checklist by various interested specialists.

CONCLUSION

An original methodical approach and checklist can improve the quality, transparency and reproducibility of scientific reports in machine learning for healthcare. The is distinguished by balanced technological and clinical aspects, educational significance and comprehensive capabilities. Independent researchers can validate the proposed approach.