Distinguishing Parkinson's disease from atypical parkinsonian syndromes using PET data and a computer system based on support vector machines and Bayesian networks
MetadataShow full item record
AuthorSegovia Román, Fermín; Álvarez Illán, Ignacio; Gorriz Sáez, Juan Manuel; Ramírez Pérez De Inestrosa, Javier; Rominger, Axel; Levin, Johannes
Frontiers Research Foundation
Bayesian networkSupport vector machine18F-DMFP PETParkinson's diseaseMultivariate analysis
Segovia, F.; et al. Distinguishing Parkinson's disease from atypical parkinsonian syndromes using PET data and a computer system based on support vector machines and Bayesian networks. Frontiers in Computational Neuroscience, 9: 137 (2015). [http://hdl.handle.net/10481/39149]
SponsorshipThis work is part of a project approved by the Andalucía Talent Hub Program launched by the Andalusian Knowledge Agency, co-funded by the European Union's Seventh Framework Program, Marie Sklodowska-Curie actions (COFUND Grant Agreement no 291780) and the Ministry of Economy, Innovation, Science and Employment of the Junta de Andalucía. The work was also supported by the University of Granada (Spain), the University for Munich (Germany), the MICINN (Spain) under the TEC2012–34306 project and the Consejera de Innovacin, Ciencia y Empresa (Junta de Andaluca, Spain) under the P11–TIC–7103 excellence project.
Differentiating between Parkinson's disease (PD) and atypical parkinsonian syndromes (APS) is still a challenge, specially at early stages when the patients show similar symptoms. During last years, several computer systems have been proposed in order to improve the diagnosis of PD, but their accuracy is still limited. In this work we demonstrate a full automatic computer system to assist the diagnosis of PD using 18F-DMFP PET data. First, a few regions of interest are selected by means of a two-sample t-test. The accuracy of the selected regions to separate PD from APS patients is then computed using a support vector machine classifier. The accuracy values are finally used to train a Bayesian network that can be used to predict the class of new unseen data. This methodology was evaluated using a database with 87 neuroimages, achieving accuracy rates over 78%. A fair comparison with other similar approaches is also provided.