Healthcare costs are rapidly increasing. As most costs are caused by inpatient medical care, there is growing interest for early detection of diseases. We believe that speech-based automated analysis has great potential for early detection of diseases. Speech carries both linguistic and paralinguistic information. Linguistic information is related to what is being said. Paralinguistic information is everything else, related to, for instance, to the physical or mental health of the speaker. Such information exists because speech production is a complex process involving a large number of organs and nerves. Disorders in these organs affect acoustical characteristics of speech. Therefore, speech provides a non-invasive, cost-effective, and scalable way to detect diseases, which is exactly what the industry is looking for. Non-invasiveness makes it easy to use speech-based analysis in any environment, which is a major benefit compared to traditional methods like blood tests and MRI. Cost-effectiveness enables applying this technology in early screening. The inherent scalability is related to two aspects: the methodology can be used to screen large populations and is also scalable to multiple disorders. The strength of speech is related to the methodology’s non-invasiveness, scalability and cost-effectiveness, which make this technology ideal for early detection of many diseases. Our initial focus is the detection of dysarthria, which is a symptom of various neurological conditions including, for example, Parkinson’s disease and strokes. The system that will be developed during this project can later be adapted to various other, different diseases such as heart failure and Alzheimer’s disease. Therefore, our long-term vision is to extend the offering to include the symptoms of other diseases as well.
