Towards an early diagnosis of Parkinson’s disease from the speech
The study of new non-invasive biomarkers able to identify the Parkinson’s disease from the speech of patients is the joint work carried out by researchers from Universidad Politécnica de Madrid, Massachusetts Institute of Technology and Johns Hopkins University.
Researchers who have a background in voice and speech technologies from Universidad Politécnica de Madrid (UPM), Massachusetts Institute of Technology (MIT) and Johns Hopkins University (JHU) are working together to design new biomarkers based on digital signal processing and machine learning for a differential diagnosis of Parkinson's disease.
The previous works, which have been recently published in PLOS ONE journal and Applied Soft Computing journal, open the door to the development of automatic screening systems and objective evaluation of the disease what it would have a great social and economic impact.
The average time to obtain a diagnosis of Parkinson’s disease is 2.9 years and is essentially based on the clinical suspicion. The diagnostic accuracy varies considerably according to the duration of the disease, age, medical expertise, and evolution. The uncertainty in the diagnosis, along with the degeneration caused before starting any treatment, has an obvious impact on the quality of life of patients.
Juan Ignacio Godino, a UPM researcher, says, “the early detection of Parkinson along with the anticipation of the start of treatment would have a relevant effect on both the quality of life of patients and the healthcare system. This would allow us to develop new therapies and better understand the disease and its evolution”.
Although robust and non-invasive early markers are not yet available, the literature has identified for decades that voice and speech are affected on the presymptomatic stages of the disease. These findings have not yet been exploited to develop reliable automatic systems for differential diagnosis and screening.
To this end, researchers from diverse centers from diverse institutions: Bioengineering and Optoelectronics Lab (UPM), Research Laboratory of Electronics (MIT) y el Center for Language and Speech Processing (JHU) are developing biomarkers based on digital signal processing and machine learning that allow an early detection, characterization, and monitoring of diverse types of neurological disorders revealed through voice, especially in Parkinson.
This study shows that the speech is a carrier of information relevant to the differential diagnosis of Parkinson's disease and that the extraction of interest features can be easily automated by assessing the different aspects related to the voice cinematic. Results give evidence that diagnostic reliability is similar to the current studies based on clinical suspicion.
This work is part of a research line that is still ongoing and requires the collaboration of citizens. The profiles are people aged between 45 and 90 who neither suffer from Parkinson’s disease nor have family history of this disease. Anyone interested in this study can fill out a contact form and the researchers responsible for the project will contact them back.
Moro-Velázquez, L.; Gómez-García, J.A.; Godino-Llorente, J.I.; Villalba, J.; Orozco-Arroyave, J.R.; Dehak, N. Analysis of speaker recognition methodologies and the influence of kinetic changes to automatically detect Parkinson's Disease. Applied Soft Computing. Vol 62. pp. 649-666. 2018.
Godino-Llorente, J.I.; Shattuck-Hufnagel, S.; Choi, J.Y.; Moro-Velázquez, L.; Gómez-García, J.A. (2017) Towards the identification of Idiopathic Parkinson’s Disease from the speech. New articulatory kinetic biomarkers. PLoS ONE 12(12): e0189583.