Quick Buzz Feed
9 min read | From predicting symptom onset to tracking movement changes via smartwatch, artificial intelligence tools are being used in research. Here’s where we are, and why Huntington’s disease is a strong candidate for these approaches.
Artificial Intelligence (AI) is designed to perform tasks typically requiring human intelligence, like language understanding or face recognition. It operates by learning patterns to make rapid, intelligent predictions. Older AI used predefined rules, while Machine Learning (ML) models create their own rules from data. Deep Learning (DL) is a more complex ML form with multiple layers, capable of finding patterns in unstructured data such as images and text.
AI offers significant advantages in healthcare, particularly for Huntington’s disease (HD) and other neurodegenerative disorders. It can provide more accessible care by reducing the need for frequent hospital visits, especially for patients in later disease stages or remote locations, and make medical care more financially sustainable. AI tools can process data, like from wearables, for motor assessments, improving convenience for patients and caregivers.
Currently, AI research for Huntington's disease (HD) focuses on modeling disease onset and progression, and serving as a diagnostic tool. Key applications include identifying genetic modifiers that influence disease onset, improving the accuracy of clinical trial recruitment, and tracking subtle movement changes through wearable devices.
A recent study utilized AI to analyze genetic data from 9,000 HD patients, aiming to understand why individuals with the same CAG repeat numbers experience different ages of disease onset. The AI models identified genetic 'modifiers'—genes beyond the primary disease gene that influence onset age—some of which were not detected in previous analyses. This research also suggested that different genes might modify onset age based on the number of CAG repeats, paving the way for more personalized HD treatment plans.
Another study employed AI to enhance recruitment for HD clinical trials by more accurately predicting the timing of symptom onset. By training an AI model with brain scan data and cognitive/motor assessment scores from natural history studies (e.g., PREDICT-HD, TRACK-HD), researchers improved prediction accuracy by 24% compared to prior methods. This capability is critical for unbiased recruitment of pre-symptomatic individuals, increasing the statistical power and reliability of trial results.
Multiple studies are exploring the use of data from wearable devices, such as smartwatches and cellphones, to monitor movement variations in individuals with HD. One study developed an AI model to differentiate involuntary HD movements (chorea) from voluntary movements, providing clinicians with more accurate insights into disease progression. Another study analyzed publicly available walking pattern data, utilizing various AI models to diagnose HD based on stride, swing, and stance intervals, with some models achieving over 80% accuracy in prediction.
Despite the significant potential of AI in healthcare, its widespread adoption is currently hindered by the 'black box' nature of the most advanced learning models, which cannot explain their conclusions. Given the critical stakes in medical care, the AI community is actively working on developing interpretable and explanatory models to build trust and facilitate broader clinical implementation.
The Huntington's disease community plays a crucial role in advancing AI-based tools through active participation in natural history studies like Enroll-HD, PREDICT-HD, and TRACK-HD. This engagement generates vast amounts of high-quality, well-organized data, which is essential for training effective AI models. The success of HD-trained AI models in improving disease prediction, personalizing prognoses, and identifying new biomarkers highlights the impact of community involvement, with ongoing efforts focused on developing more interpretable AI models for broader diagnostic and prognostic use.
Artificial intelligence (AI) is actively being applied in Huntington’s disease (HD) research as a diagnostic and monitoring tool, leveraging extensive datasets built by the HD community. AI has successfully identified previously missed genetic modifiers influencing symptom onset and improved the prediction of disease onset for clinical trial recruitment by 24%. Wearable technology paired with AI is also being used to track HD-related movement changes. A key challenge remains the lack of interpretability in advanced AI models, but the field is working on solutions. The strong participation of the HD community in natural history studies is a vital asset, providing high-quality data that drives these AI advancements.