Modern Machine-Learning Predictive Models for Diagnosing Infectious Diseases.

Area of Science:

• Computational biology
• Epidemiology
• Artificial Intelligence in Medicine

Background:

• Infectious diseases pose significant global health risks, necessitating early detection to prevent epidemics and pandemics.
• Machine learning (ML) offers promising tools for diagnosing infectious diseases at early stages.
• Existing research highlights the need for robust ML models in disease diagnostics.

Purpose of the Study:

• To review and analyze recent machine learning algorithms applied to infectious disease diagnosis.
• To identify the strengths and weaknesses of current ML models in this domain.
• To provide recommendations for advancing ML-based infectious disease diagnostic studies.

Main Methods:

• Systematic literature review of research articles from 2015 to 2022.
• Searches conducted across major scientific databases: Web of Science, ScienceDirect, PubMed, Springer, and IEEE.
• Analysis of identified ML models, focusing on their pros, cons, and data utilization.

Main Results:

• Most reviewed studies utilized small datasets, with limited use of real-time data.
• The choice of ML technique is contingent upon dataset characteristics and diagnostic objectives.
• Heterogeneous data enhances model generalization; large datasets, numerous features, and hybrid models improve performance.

Conclusions:

• Machine learning shows significant potential for infectious disease diagnosis.
• Future research should focus on utilizing large, heterogeneous, and real-time datasets.
• Integrating deep learning and natural language processing (NLP) can significantly enhance diagnostic model performance by leveraging unstructured data.