In new research published in Lancet Digital Health A new artificial intelligence (AI) model is described that groups together the typical symptom patterns of Parkinson’s disease. What stands out the most is that it also predicts the progression of these symptoms in terms of time and severity, learning from the longitudinal data of the patient.
In this sense, it is often believed that this disease is exclusive to old age but it is not. One of the most popular cases in the world is that of Michael J. Fox, the actor who played Marty McFly in the iconic films “Back to the Future.” In 1998, he announced that he had Parkinson’s and was diagnosed when he was 29 years old.
Thereafter the Fox company launched the Michael J. Fox Parkinson Research Foundation (MJFF). Its goal is to help find treatments and a cure for this disease that is estimated to affect more than six million people around the world.
Recent work on the subject
Since then, MJFF’s team of neuroscientists and strategists has worked closely with science and technology researchers, clinicians, industry allies, and patients around the world to fund the most promising research to understand and find better treatments for the illness. In July 2018, the Foundation and IBM Research announced a unique alliance with the goal of applying machine learning to advance further scientific advancements.
This collaboration reached an important milestone. In the last job published A new AI model that groups the typical symptom patterns of Parkinson’s disease is detailed by the IBM team together with MJFF scientists at Lancet Digital Health. The model also predicts the progression of these symptoms in terms of time and severity, learning from what is known as longitudinal patient data, that is, descriptions of a patient’s clinical status collected over time.
The goal is to use AI to assist in the management and design of clinical trials. These goals are important because, despite the prevalence of Parkinson’s, patients experience a unique variety of symptoms, both motor and non-motor.
The use of machine learning to learn from vast amounts of patient data enables clinicians and researchers to have a new tool to better predict the markedly variable progression of symptoms in individual Parkinson’s patients. Likewise, that this allows to manage and treat the disease more effectively, and that it gives rise to the possibility of identifying the best candidates for clinical trials that are more specific and effective.
Putting AI to work
The results are the next step in previously published research. That work focused on developing a method for some of the unique challenges of healthcare applications, including enabling personalized predictions and accounting for the effects of drugs on symptom measurements.
New insights into disease progression
These modeling decisions have allowed researchers to obtain more information about disease states and pathways of progression. The results suggest that a patient’s condition may vary in a number of factors, such as the ability to perform daily activities; problems related to motor slowness, tremor, and postural instability; as well as non-motor symptoms, including depression, anxiety, cognitive impairment, and sleep disorders.
The results support the hypothesis that there are various pathways of progression, as indicated by the many disease trajectories that have been studied. However, the AI model can still make accurate predictions. Because the model draws from a data set, it has been able to successfully predict an advanced stage of Parkinson’s disease associated with outcomes such as dementia and the inability to walk without assistance.
Due to the diversity of experiences in Parkinson’s disease, it is hoped that by allowing such predictions, the model can help with management and provide more specific inclusion and outcome definition criteria during clinical trial design.