Researchers teach AI how to diagnose severity of ALS

This breakthrough could lay the foundation for a new method of treating ALS.

TORONTO–Scientists from the KITE Research Institute taught artificial intelligence how to determine the severity of a person’s amyotrophic lateral sclerosis (ALS) by analyzing speech patterns. 

ALS is a fatal progressive neurodegenerative disease that causes patients to lose control of muscles responsible for speaking, eating, and breathing. Currently there is no cure for this disease.

Treatments for ALS are most effective in the early stages. Unfortunately, this disease is difficult to diagnose early on as many of its symptoms are similar to other illnesses. 

This breakthrough could lay the foundation for a new method of diagnosing ALS. 

The results of the team’s research were published in BioMedical Engineering OnLine.

The International Conference on Aging, Innovation & Rehabilitation (ICAIR) partnered with BioMedical Engineering OnLine to publish a special edition of the journal featuring full-length papers of the highest scoring abstracts from the conference.

The ICAIR organizing committee spoke with the paper’s first author KITE trainee Dr. Leif Simmatis to learn more about this exciting work. 

Which patient groups are most affected by this? 

This study was performed with data from amyotrophic lateral sclerosis (ALS) patients, and so the findings are most directly-applicable to that population. However, the methodology and approach to data analysis could potentially apply to any other disease where speech and voice functions are well known to be affected (e.g., Parkinson’s disease).

 What did you find?

Using an automated speech analysis system to extract various acoustic measures of speech function, we were able to differentiate between different ALS severities, as well as between ALS and healthy controls. We additionally identified that well-established measures of speech function were important for differentiating between groups. 

 Why does this matter? 

Speech is of great interest for assessing neurodegenerative diseases such as ALS, and speech data can be easily collected e.g., from patients‘ homes. By applying automated speech analysis pipelines and interpretable machine learning methods, we can gain important insights into disease processes, and develop methods for screening and stratifying patients. 

 What is the potential impact? 

This work provided an important demonstration of the ability of automated speech-based analysis systems to capture disease-related phenomena in ALS. By identifying objective measures of speech function, it may be possible to stratify patients for trials of novel therapeutics and track their effects over time, which may help alleviate some established issues in ALS clinical trials.

Research Spotlight: 
Using automatic acoustic analysis to detect ALS

Postdoctoral Fellow at the KITE Research Institute

Name of Publication:
Detecting bulbar amyotrophic lateral sclerosis (ALS) using automatic acoustic analysis

Name of Journal:
BioMedical Engineering OnLine