Using Machine Learning to Detect Parkinson’s Disease in Patients’ Voices
Researchers are continuously exploring new methods to detect neurodegenerative diseases early, and a recent study has focused on using machine learning models to ‘hear’ Parkinson’s disease in patients’ voices.
According to bioinformatician Aniruth Ananthanarayanan and his team at the University of North Texas, their findings suggest that voice-based machine learning models can identify disease signatures even before visible motor signs appear in patients. This research is yet to be published but holds promise for early detection of Parkinson’s disease.
Parkinson’s disease affects nearly 9 million people globally and is characterized by difficulties in controlling fine movements, tremors in extremities, as well as challenges in mood, thinking, and memory. While the mechanisms of the disease are somewhat understood, the triggers for its onset are still not fully comprehended.
There is currently no cure for Parkinson’s disease, making early detection crucial for implementing therapies that can slow down symptoms and improve patients’ quality of life. Ananthanarayanan and his team utilized machine learning models to differentiate between volunteers with and without Parkinson’s disease based on their voice recordings.
The study involved training the models on 195 voice recordings from 31 individuals, 23 of whom were diagnosed with Parkinson’s. The machine learning program accurately identified patients with the condition in 90% of its attempts. Vocal features such as jitter, noise-to-harmonic ratios, and disordered voice signal patterns were assessed, all of which have been linked to Parkinson’s symptoms.
While the study shows promise for early screening of Parkinson’s disease through voice analysis, further research is needed to validate the models’ generalization across different populations. Data scientist Aiden Arnold commented on the study, stating that the voice-based approach holds potential as an early screening tool.
If the findings from this research hold true across wider populations, a voice-based screening tool could be a scalable and affordable option for early detection as Parkinson’s disease cases continue to rise. The study is currently awaiting peer review and has been uploaded to MedRxiv for further scrutiny.
