Machine learning in neurodegenerative disease research: a growing trend

By Dr. Sanchari Sinha Dutta, Ph.D.Aug 4 2023Reviewed by Benedette Cuffari, M.Sc.

In a recent review posted to the medRxiv* preprint server, scientists explore the utility of machine learning methods in the field of neurodegenerative disease diagnosis, prognosis, and treatment effect prediction.

Study: The use of machine learning methods in neurodegenerative disease research: A scoping review. Image Credit: sfam_photo / Shutterstock.com

*Important notice: medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.

Background

Neurodegenerative diseases are detrimental age-related pathological conditions associated with progressive deterioration of the neuronal network in the central and peripheral nervous systems. As a result, all neurodegenerative diseases are associated with progressively disabling symptoms that ultimately lead to complete loss of autonomy and death.

The most common neurogenerative diseases include Alzheimer’s disease, Parkinson’s disease, multiple sclerosis, amyotrophic lateral sclerosis, and Huntington’s disease.

In the United States, both Alzheimer’s disease and Parkinson’s disease are the most common neurodegenerative diseases. Current estimates indicate that up to 6.2 million people are living with Alzheimer’s disease in the U.S., whereas Parkinson’s disease currently affects about one million Americans. As life expectancy increases in many nations throughout the world, researchers predict that the prevalence of these neurodegenerative diseases will also rise.

To improve the management of these incurable diseases, it is important to understand disease pathogenesis, develop accurate diagnostic and prognostic tools, and discover targeted therapies. The use of machine learning methods is increasing in the field of neurodegenerative disease research for rapidly and accurately analyzing disease-related data, which is essential for supporting diagnostic and therapeutic innovations.

In the current scoping review, scientists explore the utility of machine learning methods in the study of the five most prevalent neurodegenerative diseases, including Alzheimer’s disease, multiple sclerosis, amyotrophic lateral sclerosis, Parkinson’s disease, and Huntington’s disease.

Study design

Various scientific databases were searched to identify studies that utilized machine learning methods for the diagnosis, prognosis, and treatment prediction of five neurodegenerative diseases. All studies published between January 2016 and December 2020 were included in the analysis.

A total of 4,471 studies were screened, 1,485 of which were ultimately included in the final analysis. The information extracted from each study included type of neurodegenerative disease, publication year, sample size, machine learning algorithm data type, primary clinical goal, and machine learning method type. Both qualitative and quantitative analyses of the study results were conducted.   

The growing use of machine learning methods

A gradual increase in the use of machine learning methods in neurodegenerative diseases was observed over time. More specifically, the number of studies using these methods increased from 172 in 2016 to 490 in 2020, thus reflecting a 185% increase in the incorporation of this technology. Alzheimer’s disease and Parkinson’s disease were the most studied neurodegenerative diseases using machine learning methods.

In the selected studies, imaging was the most commonly analyzed data type, followed by functional, clinical, biospecimen, genetic, electrophysiological, and molecular analyses. Imaging and functional data were the most commonly used data types in Alzheimer’s disease and Parkinson’s disease, respectively. About 68% of imaging data was related to Alzheimer’s disease and 76% of functional data was related to Parkinson’s disease.

Regarding primary clinical goals, machine learning methods were most frequently used for disease diagnosis, followed by disease prognosis and prediction of treatment effects. Imaging data remained the most commonly used data type for disease diagnosis and prognosis. For the prediction of treatment effect, functional data were the most commonly used data type.

A total of 2,734 types of machine learning methods were used in the selected studies. Among these methods, support vector machine, random forest, and convolutional neural network were most frequently noted. In addition, 322 unique methods were identified in the review.  

Significance

The current scoping review indicates an increase in the application of machine learning methods in neurodegenerative disease research. The scientists explain that the popularity of these methods is increasing to improve the clinical course of these detrimental diseases.

Although certain treatments are currently available to alleviate some of the physical and mental symptoms associated with neurodegenerative diseases, there remains a lack of therapies capable of slowing the progression of neuronal death. Thus, there remains an urgent need to increase the application of machine learning methods to identify prognostic biomarkers and discover novel therapeutics for the treatment of neurodegenerative diseases.

*Important notice: medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.