Portuguese has created a website that detects Parkinson’s disease through voice


This is a site that analyzes voice files, and thus determines if the user has evidence of Parkinson’s disease, thus allowing early detection and treatment, the site was developed by an engineer of the Institute of Engineering of Porto (ISEP).

In addition to allowing the early detection of Parkinson’s disease (in patients of different ages), this technology also aims to reduce the costs associated with exams and treatments, explained Diogo Braga, the person in charge of creating this project. Fiogo Braga recently completed a degree in Computer Engineering at ISEP.

“There is a high cost associated with neurodegenerative diseases, with costly treatments that only serve to give a better quality of life than patients would have without treatments, which have to be scrupulously followed to take effect”, the engineer explained.

But, according to Diogo Braga, this disease has other costs besides treatment, which mainly concern the caregivers, the depressions patients develop and the fact that they do not have activity in the labor market.

“When a sum of these costs is made, the result is quite high”, said Diogo Braga, adding that early treatments not only lead to a reduction of associated costs, but also to the progression of the disease to certain stages.

In order to obtain the result, it is enough that users (anonymous or identified) download on the site an audio file (in ‘wav’ or ‘mp3’ format) with about five seconds. The result is then immediately reported, telling the user if you have any clues, whether to see a doctor or to repeat the test on another day.

These indications are detected through a machine-learning system based on data from 24 patients with Parkinson’s at different stages of disease progression, as well as data from 30 healthy people, 92.38% during the tests.

Diogo Braga explained that Parkinson’s disease can be detected by voice stiffness and vowel cycles (which cannot be perceived by humans).

“When we express a vowel our mouth cycles, which is supposed to be fluid, without any rigidity and without braking,” he said, further clarifying that in Parkinson’s patients, it is possible to detect these factors in the voice at an early stage of the disease.

For the engineer, the differentiating factor of this technology is that the detection is made “with ambient noise tolerance”.

“Because users may not be in lab conditions, it is important that the system can detect signs of the disease through the voice even when the recording is light”, he said.

Developed within the scope of the curriculum internship that the recently graduated from ISEP carried out in the Group of Research in Engineering and Intelligent Computing for Innovation and Development (GECAD), the site took about four months to be developed.

The platform, which may be adapted in the future for other neurodegenerative diseases, such as Alzheimer’s and Huntington, is not yet online at the moment.

The ISEP teachers and project supervisors, Ana Maria Madureira and Luís Filipe Coelho, point out that priority has been given to the publication of scientific articles, which allow “to validate and give a greater degree of credibility to the project”.