1. Comparison between K-nearest-neighbor approaches for conditional entropy estimation: Application to the assessment of the cardiac control in amyotrophic lateral sclerosis patients
- Author
-
Andrea Marchi, Beatrice De Maria, Vlasta Bari, Laura Adelaide Dalla Vecchia, Gabriele Mora, Riccardo Sideri, Alberto Porta, and Kalliopi Marinou
- Subjects
Adult ,Male ,Computer science ,Entropy ,Rest ,0206 medical engineering ,02 engineering and technology ,Machine learning ,computer.software_genre ,k-nearest neighbors algorithm ,03 medical and health sciences ,Entropy (classical thermodynamics) ,0302 clinical medicine ,Heart Rate ,Tilt-Table Test ,medicine ,Cluster Analysis ,Humans ,Entropy (information theory) ,Entropy (energy dispersal) ,Amyotrophic lateral sclerosis ,Entropy (arrow of time) ,Aged ,Aged, 80 and over ,Conditional entropy ,medicine.diagnostic_test ,business.industry ,Entropy (statistical thermodynamics) ,Amyotrophic Lateral Sclerosis ,Heart ,Pattern recognition ,Mutual information ,Middle Aged ,medicine.disease ,020601 biomedical engineering ,Linear Models ,Female ,Artificial intelligence ,business ,computer ,Random variable ,Electrocardiography ,Algorithms ,030217 neurology & neurosurgery ,Entropy (order and disorder) - Abstract
The study evaluates the k-nearest-neighbor (KNN) strategy for the assessment of complexity of the cardiac neural control from spontaneous fluctuations of heart period (HP). Two different procedures were assessed: i) the KNN estimation of the conditional entropy (CE) proposed by Porta et al; ii) the KNN estimation of mutual information proposed by Kozachenko-Leonenko, refined by Kraskov-Stögbauer-Grassberger and here adapted for the CE estimation. The two procedures were compared over HP variability recordings obtained at rest in supine position and during head-up tilt (HUT) in amyotrophic lateral sclerosis patients and healthy subjects. We found that the indexes derived from the two procedures were significantly correlated and both methods were able to detect the effect of HUT on HP complexity within the same group and distinguish the two populations within the same experimental condition. We recommend the use of the KNN strategy to quantify the dynamical complexity of cardiac neural control in addition to more traditional approaches.
- Published
- 2016