Classical FE Analysis to Classify Parkinson’s Disease Patients
Parkinson’s disease (PD) is a neurodegenerative condition that affects the correct functioning of the motor system in the human body. Patients exhibit a reduced capability to produce facial expressions (FEs) among different symptoms, namely hypomimia. Being a disease so hard to be detected in its ea...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/11/21/3533 |
| _version_ | 1797468525878575104 |
|---|---|
| author | Nestor Rafael Calvo-Ariza Luis Felipe Gómez-Gómez Juan Rafael Orozco-Arroyave |
| author_facet | Nestor Rafael Calvo-Ariza Luis Felipe Gómez-Gómez Juan Rafael Orozco-Arroyave |
| author_sort | Nestor Rafael Calvo-Ariza |
| collection | DOAJ |
| description | Parkinson’s disease (PD) is a neurodegenerative condition that affects the correct functioning of the motor system in the human body. Patients exhibit a reduced capability to produce facial expressions (FEs) among different symptoms, namely hypomimia. Being a disease so hard to be detected in its early stages, automatic systems can be created to help physicians in assessing and screening patients using basic bio-markers. In this paper, we present several experiments where features are extracted from images of FEs produced by PD patients and healthy controls. Classical machine learning methods such as local binary patterns and histograms of oriented gradients are used to model the images. Similarly, a well-known classification method, namely support vector machine is used for the discrimination between PD patients and healthy subjects. The most informative regions of the faces are found with a principal component analysis algorithm. Three different FEs were modeled: angry, happy, and surprise. Good results were obtained in most of the cases; however, happiness was the one that yielded better results, with accuracies of up to 80.4%. The methods used in this paper are classical and well-known by the research community; however, their main advantage is that they provide clear interpretability, which is valuable for many researchers and especially for clinicians. This work can be considered as a good baseline such that motivates other researchers to propose new methodologies that yield better results while keep the characteristic of providing interpretability. |
| first_indexed | 2024-03-09T19:08:39Z |
| format | Article |
| id | doaj.art-b13b84bfa40445b6b9fa4907d7d27ebd |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T19:08:39Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-b13b84bfa40445b6b9fa4907d7d27ebd2023-11-24T04:25:24ZengMDPI AGElectronics2079-92922022-10-011121353310.3390/electronics11213533Classical FE Analysis to Classify Parkinson’s Disease PatientsNestor Rafael Calvo-Ariza0Luis Felipe Gómez-Gómez1Juan Rafael Orozco-Arroyave2GITA Lab, Electronics and Telecommunications Department, Faculty of Engineering, Universidad de Antioquia, Medellín 050010, ColombiaGITA Lab, Electronics and Telecommunications Department, Faculty of Engineering, Universidad de Antioquia, Medellín 050010, ColombiaGITA Lab, Electronics and Telecommunications Department, Faculty of Engineering, Universidad de Antioquia, Medellín 050010, ColombiaParkinson’s disease (PD) is a neurodegenerative condition that affects the correct functioning of the motor system in the human body. Patients exhibit a reduced capability to produce facial expressions (FEs) among different symptoms, namely hypomimia. Being a disease so hard to be detected in its early stages, automatic systems can be created to help physicians in assessing and screening patients using basic bio-markers. In this paper, we present several experiments where features are extracted from images of FEs produced by PD patients and healthy controls. Classical machine learning methods such as local binary patterns and histograms of oriented gradients are used to model the images. Similarly, a well-known classification method, namely support vector machine is used for the discrimination between PD patients and healthy subjects. The most informative regions of the faces are found with a principal component analysis algorithm. Three different FEs were modeled: angry, happy, and surprise. Good results were obtained in most of the cases; however, happiness was the one that yielded better results, with accuracies of up to 80.4%. The methods used in this paper are classical and well-known by the research community; however, their main advantage is that they provide clear interpretability, which is valuable for many researchers and especially for clinicians. This work can be considered as a good baseline such that motivates other researchers to propose new methodologies that yield better results while keep the characteristic of providing interpretability.https://www.mdpi.com/2079-9292/11/21/3533Parkinson’s Diseaseimage processinghypomimiaFEclassic techniquesmachine learning |
| spellingShingle | Nestor Rafael Calvo-Ariza Luis Felipe Gómez-Gómez Juan Rafael Orozco-Arroyave Classical FE Analysis to Classify Parkinson’s Disease Patients Electronics Parkinson’s Disease image processing hypomimia FE classic techniques machine learning |
| title | Classical FE Analysis to Classify Parkinson’s Disease Patients |
| title_full | Classical FE Analysis to Classify Parkinson’s Disease Patients |
| title_fullStr | Classical FE Analysis to Classify Parkinson’s Disease Patients |
| title_full_unstemmed | Classical FE Analysis to Classify Parkinson’s Disease Patients |
| title_short | Classical FE Analysis to Classify Parkinson’s Disease Patients |
| title_sort | classical fe analysis to classify parkinson s disease patients |
| topic | Parkinson’s Disease image processing hypomimia FE classic techniques machine learning |
| url | https://www.mdpi.com/2079-9292/11/21/3533 |
| work_keys_str_mv | AT nestorrafaelcalvoariza classicalfeanalysistoclassifyparkinsonsdiseasepatients AT luisfelipegomezgomez classicalfeanalysistoclassifyparkinsonsdiseasepatients AT juanrafaelorozcoarroyave classicalfeanalysistoclassifyparkinsonsdiseasepatients |