During the past three years, artificial intelligence (AI) researchers at IBM in Armonk have collaborated with neuroscientists at the Michael J. Fox Foundation on studies designed to predict the progression of Parkinson”™s disease.
In a study that was recently published in The Lancet Digital Health, the IBM and Fox foundation team showed the ability to build AI in a manner that could accurately chart the patterns in which a patient”™s Parkinson”™s symptoms progress, as well as how and when a patient”™s health would devolve into an acute state of the disease.
Jianying Hu, IBM fellow and global science leader for AI in health care, explained that trying to track how Parkinson”™s progresses within a patient is one of the most challenging tasks facing the medical profession.
“It”™s a neurodegenerative condition that has no cure,” she said. “It affects 6 million people and the case rate is rising rapidly. At the same time, there is very little understanding of how the disease actually progresses ”” and its progression is very heterogeneous, meaning different patients can take a completely different route.
“That light lack of understanding into the disease progression has really been hampering our ability to accurately assess where the patient is at and to be able to appropriately treat and manage patients,” she said. “And, also, to drive these types of clinical trials.”
Hu praised the foundation for amassing a large collection of observational data on Parkinson”™s disease through an international study.
Using that data as a foundation, Hu continued, the next obstacle was to work through the heterogeneity and complexity of the disease to find common ground among patients.
“The disease progresses along many dimensions,” Hu said. “There is the motor dimension, which most people are aware of and also nonmotor symptoms like the disorders and mood swings. For that reason, the effect of the medications that used to manage the symptoms vary from patient to patient and also depends on the stage of the disease.”
As a result, traditional disease modeling did not apply to Parkinson”™s. Hu noted the foundation contacted IBM to leverage its machine learning and AI methodologies “to help them tease out useful insights from that data.”
By using IBM”™s AI tools, the researchers”™ modeling decisions opened a new door into understanding the disease”™s multiple states and progression pathways.
The results have determined the factors that define a patient”™s state, including the ability to perform routine daily activities, issues concerning degrading of movement and postural instability and nonmotor symptoms, including depression, anxiety, cognitive impairment and sleep disorders.
While the researchers identified diverse progression pathways, the AI model was still able to make accurate predictions, including forecasts into the advanced state of Parkinson”™s disease and outcomes, including dementia and the inability to walk unassisted.
Hu also pointed out that IBM”™s AI researchers have been working with other research foundations to use their technology to study the progression of Huntington”™s disease and type 1 diabetes, with the goal of using their research to enhance clinical trials and accelerate the development of new therapeutics.
“This is really, really exciting,” Hu said. “We have been working, applying and developing advanced machine learning methods to observational health care data to build models and derive insights.
“And the key challenge is to be able through that collaboration to really understand what is the right problem to solve.”
