Applications of machine learning to diagnosis and treatment of neurodegenerative diseases

Machine learning is a branch of artificial intelligence increasingly applied in clinical research and drug development. It can help researchers in brain health to recognize patterns in large, multidimensional data sets that help generate clinical and research insights across the course of health and disease. This can be done using imaging information, genetic and molecular biomarkers, risk factors and/or treatment responses.

A new review from academics at the University of Sheffield and the Benevolent AI drug discovery company summarizes the types of machine learning available to brain health researchers and outlines their potential applications.1 Task-driven activities based on labelled datasets (supervised), pattern identification and recognition (unsupervised) and reward-driven repetition (reinforcement) are the three main categories of machine learning. Each is suited to particular research outcomes:

  • Diagnosis and prognosis: relevant data sources include neuroimaging, motor function, language features and molecular and genetic data.
  • Therapy development: potential uses include target identification and patient stratification based on diagnostic and prognostic factors.

Reference

  1. Myszczynska MA et al. Applications of machine learning to diagnosis and treatment of neurodegenerative diseases. Nat Rev Neurol 2020;16:440–56. https://www.nature.com/articles/s41582-020-0377-8