Devendra Kumar
Researchers from the Francis Crick Institute and the UCL Queen Square Institute of Neurology in London have developed an artificial intelligence (AI) tool that can classify four subtypes of Parkinson’s disease with up to 95 percent accuracy. The team “trained” a computer program to recognize the subtypes of the condition using images of stem cells from patients. This breakthrough could pave the way for personalized medicine and targeted drug discovery.
The researchers understand many of the processes that cause Parkinson’s in people’s brains, but they have no way of knowing which mechanism is occurring while the patient is still alive. This lack of knowledge makes it difficult to provide precise treatments. The current treatments for Parkinson’s do not make a significant difference in the progression of the disease. The AI tool uses a model of the patient’s own neurons and combines it with large numbers of images to generate an algorithm that can classify certain subtypes.
The researchers believe that their platform could allow them to test drugs in stem cell models and predict whether a patient’s brain cells would be likely to respond to a drug before enrolling in clinical trials. This could lead to fundamental changes in how personalized medicine is delivered. Parkinson’s is a condition in which parts of the brain become progressively damaged over many years, causing a range of physical and psychological symptoms that vary from person to person due to differences in the underlying mechanisms.
Until now, there has been no way to accurately differentiate Parkinson’s subtypes, which means people are given nonspecific diagnoses and do not always have access to targeted treatments, support, or care. To develop the AI tool, the researchers generated stem cells from patients’ own cells and used those cells to create four different subtypes of Parkinson’s. They then worked with the British technology company Faculty AI to develop machine-learning algorithms that could accurately predict the Parkinson’s subtype when presented with images it had not seen before.
The team’s approach enabled them to evaluate a larger number of cell features and assess the importance of these features in discerning the disease subtype. By using deep learning, they were able to extract more information from their images than with conventional image analysis. The researchers hope to expand this approach to understand how these cellular mechanisms contribute to other subtypes of Parkinson’s.