Animo

Animo
GIF

Inspiration

Given our team’s medical backgrounds, many of us have witnessed first hand the devastation that neurodegenerative diseases can cause first-hand. We understand first hand the many challenges that these patients face - from long waiting lists to poorly personalised regimes. Given our long-standing interest in the research space of these diseases, combined with our technical backgrounds, we aspire to design a patient-centric technological solution to better connect clinicians with the well-being of their patients.

What it does

Clinicians measure Parkinson’s severity using a UPDRS score to use as a basis for drug prescription. Studies have shown machine learning models capable of predicting said UPDRS score from patient vocal recordings. It is estimated that 90% of Parkinson’s patients have some form of vocal issue in early disease stages. Our app collects patient voice recording of the ‘a’ vowel phonation into the cloud system and predicts disease severity with machine learning. This will allow a better understanding of disease progression to clinicians and ultimately better drug management. Research seems to be very supportive of translating this to other neurodegenerative diseases, such as Alzheimer's.

How I built it

Our prototype machine learning model is developed from an open-source database containing 188 Parkinson’s patients and 64 healthy control patients. Using Python Scikit-learn, our decision tree classifier has achieved ROC AUC values at 0.80, with a specificity of 0.64 and sensitivity of 0.9. The 100 best features were selected using SelectKBest. This algorithm runs on feature-extracted audio waveforms, collected from our phone app.

Challenges I ran into

The database mentioned above does not have patient severity measurements. We hope to extend our prototype algorithm with a more representative database and advanced feature extraction and analysis, to improve its sensitivity and specificity. We would also demonstrate this in a clinical trial, as required by FDA guidelines.

Accomplishments that I'm proud of

Speech analysis for Parkinson’s severity prediction is a new and rapidly growing field with great potential in improving patient drug management, and our solution at this time appears to be novel. The opportunity to equip clinicians with daily data of disease severity will clearly benefit the quality of life of Parkinson’s patients.

What I learned

Speech analysis for Parkinson’s severity prediction is a growing field with great potential in improving patient drug management. The opportunity to equip clinicians with daily data of disease severity will clearly benefit the quality of life of Parkinson’s patients.

What's next for Animo

We have been in touch with experts in the field of Parkinson research, and the feedback so far has been very positive on both the problem and the solution we’ve identified. Going forward, we hope to partner with both public health bodies and external funding initiatives to bring this vision to life.

To continue our research in the US we would hope to partner with leading organisations such as the Michael J Fox research group for Parkinson's Disease as well as applying for an NIH grant to pursue the mobileHealth space. We additionally would consider a joint venture with companies such as Novartis who manufacture levodopa.