Source: Agencia ID
Students at the School of Computing (Escom) of the National Polytechnic Institute (IPN) developed a computer system that allows diagnosing Parkinson’s disease since its initial stage by associative memories and voice analysis.
Students Alberto Cruz Cruz, Ricardo Lopez Vicente and Erika Robledo Alonso created this prototype, with which they will obtain their degree on Computer Systems Engineering.
The authors explained that Parkinson’s disease, in its early stages, is manifested by alterations in speech caused by tightness in the larynx and vocal cords. They mentioned that Parkinson’s disease is difficult to diagnose because it is often confused with common aging signs central nervous system diseases, such as multiple sclerosis.
In this regard, they talked about the need to have sufficient tools to achieve an early diagnosis and reduce side effects caused by medications used to control the condition.
Advised by Escom professors and researchers, María Julia Calderón Sambarino and José Félix Talamantes Serrano, and supported by the Mexican Association of Parkinson’s, the group of students devoted two years for researching and building the system.
For its design, they used a database donated by Oxford University to Machine Learning Repository, which contained approximately 23 voice parameters. They analysed these parameters through various associative memories. As they are extraction models of small pieces of information, they have a very high yield; thus, the error rate is minimal.
They explained that by adapting alpha-beta and morphological memories to the system models, it was possible to perform more accurate diagnoses, reaching up to one hundred percent effectiveness.
“In the repository, we had 195 records; and the memories determined that 26 percent of them did not have Parkinson’s disease, but the remaining 74 percent did, achieving a hundred percent accuracy. We proved that associative memories are models with a high effectiveness percentage, because they are very strict regarding information retrieval,” they concluded.
Source: La Jornada
