Artificial intelligence in disease diagnosis: Application within feature and state spaces

Aim: To justify strategies for transitioning from the topology of state and feature spaces to a unified, structured metric space within medical decision-making.

Materials and Methods: The paper proposes the simultaneous use of two conceptual spaces-states and features - when applying artificial intelligence in healthcare to support evidence-based diagnostic decision-making. To implement classification mechanisms using artificial intelligence, it is proposed to use the Multiscale Classifier algorithm with additional spatial analysis to extract features and evaluate the quality of the obtained classifications using ROC curves (Receiver Operating Characteristic Curves). For modeling the state space, it is effective to use recursive Bayesian state estimation, dynamic regression, and correlation analysis.

Conclusions: Findings indicate that the accuracy and efficiency of classification processes can be substantially enhanced by adopting dynamic classification principles. Moreover, the overall effectiveness of AI deployment in healthcare partially depends on the extent of generalization and the specific structural organization of medical information. A critical and practical element at the pre-implementation stage of AI integration is the development of a domain-specific ontology.