Neurodegenerative diseases can be tracked using a machine-learning method
Neurodegenerative disorders, such as amyotrophic lateral sclerosis, Alzheimer’s and Parkinson’s, are complex, chronic illnesses that present with a wide range of symptoms and worsen at varying rates. They have a number of genetic and environmental causes which, in some cases, remain unknown. ALS is a fatal disease that affects voluntary muscles. While most people only live a few months after being diagnosed, some sufferers may have the condition for many years. The symptoms of ALS are also different. A slower disease progression is often associated with limbs that have fine motor skills. Bulbar ALS, on the other hand, affects swallowing, speech, breathing and mobility. Understanding the progression of diseases such as ALS is crucial for enrolling in clinical trials, analyzing potential interventions, and discovering root causes.
It is not easy to assess the progression of a disease. Clinical studies assume that the health of patients declines in a linear fashion based on a symptom scale. These linear models are then used to determine whether drugs slow down disease progression. Data indicate, however, that ALS follows nonlinear trajectories. Symptoms can alternate between periods of stability and rapid change. Comparing patient populations is difficult because data are sparse and health assessments often depend on subjective ratings metrics that are measured at irregular time intervals. The heterogeneous data and progress complicates analyses of invention effectiveness, and may mask the disease origin.
Researchers from MIT, IBM Research and other institutions have developed a new method of machine learning to characterize ALS progression patterns in order to improve clinical trial design.