Titre : Contribution à l'étude des capacités motrices et cognitives des personnes âgées par traitement de données multivariées
Thèse de doctorat de l'université d'Angers. Ecole doctorale n°601 Mathématiques et Sciences et Technologies de l’Information et de la Communication (Spécialité : Signal, Image, Vision). Unité de recherche : Laboratoire Angevin de Recherche en Ingénierie des Systèmes (LARIS), EA 7315. Thèse CIFRE N°2017/1165
Soutenue le : Jeudi 26 novembre 2020
Jury :
| Présidente | Régine LE BOUQUIN-JOANNES | Professeur, Université de Rennes 1 |
| Examinateur et rapporteur | Pierre-Yves GUMERY | Professeur, Université de Grenoble Alpes |
| Examinateur et rapporteur | Laurent ARSAC | Professeur, Université de Bordeaux |
| Directrice | Anne HUMEAU-HEURTIER | Professeur, Université d'Angers |
| Co-Directeur | Pierre ABRAHAM | Professeur d’université - Praticien Hospitalier, Université d'Angers |
| Co-Encadrant | Cédric ANNWEILER | Professeur d’université - Praticien Hospitalier, Université d'Angers |
Mots clefs : signaux multivariés; entropie; théorie de l'information; vieillissement; réalité virtuelle; double tâche; analyse non linéaire
Résumé : Le vieillissement est un phénomène complexe ; 30% de la population française en 2020 est considérée comme âgée. Le diagnostic en gériatrie devient donc un enjeu majeur pour notre société pour prévenir et guérir les pathologies liées à l'âge. L'évolution des technologies, avec notamment la réalité virtuelle, permet de développer de nouvelles méthodes de diagnostic plus ludiques. Cycléo est un simulateur de vélo, adapté aux personnes âgées, utilisé actuellement de manière récréative dans des établissements pour seniors. Au cours de son utilisation, le vélo enregistre différentes données comportementales. Dans cette thèse, nous souhaitons répondre à la question suivante : est-il possible de transformer cet appareil de loisir en un outil d'aide au diagnostic gériatrique ? Pour ce faire, nous avons utilisé différentes méthodes de traitement du signal (entropie, corrélation croisée, entropie croisée, décomposition modale empirique, diagramme de Bland-Altman, ...) sur les données enregistrées par le vélo. Ce travail préliminaire se structure en deux parties. Dans un premier temps, nous avons considéré les données de Cycléo lors d'un usage récréatif afin d'étudier le rôle de différentes conditions d'utilisation et l'influence de l'âge sur nos résultats. Dans un second temps, nous avons intégré Cycléo dans un protocole d'essai clinique. Notre objectif était d'étudier les éventuelles correspondances entre nos résultats et ceux obtenus grâce à l'utilisation du tapis de marche couramment employé dans les diagnostics gériatriques.
Cliquer sur un article pour voir ses détails.
Abstract:
Background: Venous thromboembolism (VTE) is a common health issue. A clinical expression of VTE is a deep vein thrombosis (DVT) that may lead to pulmonary embolism (PE), a critical illness. When DVT is suspected, an ultrasound exam is performed. However, the characteristics of the clot observed on ultrasound images cannot be linked with the presence of PE. Computed tomography angiography is the gold standard to diagnose PE. Nevertheless, the latter technique is expensive and requires the use of contrast agents.
Purpose: In this article, we present an image processing method based on ultrasound images to determine whether PE is associated or not with lower limb DVT. In terms of medical equipment, this new approach (Doppler ultrasound image processing) is inexpensive and quite easy.
Methods: With the aim to help medical doctors in detecting PE, we herein propose to process ultrasound images of patients with DVT. After a first step based on histogram equalization, the analysis procedure is based on the use of bi-dimensional entropy measures. Two different algorithms are tested: the bi-dimensional dispersion entropy (DispEn2D) mesure and the bi-dimensional fuzzy entropy (FuzEn2D) mesure. Thirty-two patients (12 women and 20 men, 67.63±16.19 years old), split into two groups (16 with and 16 without PE), compose our database of around 1490 ultrasound images (split into seven different sizes from 32x32 px to 128x128 px). p-values, computed with the Mann-Whitney test, are used to determine if entropy values of the two groups are statistically significantly different. Receiver operating characteristic (ROC) curves are plotted and analyzed for the most significant cases to define if entropy values are able to discriminate the two groups.
Results: p-values show that there are statistical differences between FuzEn2D of patients with PE and patients without PE for 112x112 px and 128x128px images. Area under the ROC curve (AUC) is higher than 0.7 (threshold for a fair test) for 112x112 and 128x128 images. The best value of AUC (0.72) is obtained for 112x112 px images.
Keywords: bidimensional entropy, pulmonary embolism, ultrasound images
Lien vers l'articleAbstract: The evidence‐based medicine allows the physician to evaluate the risk‐benefit ratio of a treatment through setting and data. Risk‐based choices can be done by the doctor using different information. With the emergence of new technologies, a large amount of data is recorded offering interesting perspectives with machine learning for predictive data analytics. Machine learning is an ensemble of methods that process data to model a learning problem. Supervised machine learning algorithms consist in using annotated data to construct the model. This category allows to solve prediction data analytics problems. In this paper, we detail the use of supervised machine learning algorithms for predictive data analytics problems in medicine. In the medical field, data can be split into two categories: medical images and other data. For brevity, our review deals with any kind of medical data excluding images. In this article, we offer a discussion around four supervised machine learning approaches: information‐based, similarity‐based, probability‐based, and error‐based approaches. Each method is illustrated with detailed cardiovascular and nuclear medicine examples. Our review shows that model ensemble (ME) and support vector machine (SVM) methods are the most popular. SVM, ME, and artificial neural networks often lead to better results than those given by other algorithms. In the coming years, more studies, more data, more tools, and more methods will, for sure, be proposed.
Keywords: prediction; supervised algorithm; information-based approach; similarity-based approach; error-based approach; probability-based approach
Lien vers l'articleAbstract: Studying the impact of age is important to understand the phenomenon of aging and the disorders that are associated with it. In this work, we analyze age-related alterations on the capacities to navigate on a bike. For this purpose, we use CycléoONE, a bike simulator, and entropy measures. We thus record navigation data (handlebar angle and speed) during the ride. They are processed with two cross-distribution entropy methods (time-shift multiscale cross-distribution entropy and multiscale cross-distribution entropy). We also analyze the time series with a detrended cross-correlation analysis to determine which method can best underline age-related alterations. Our results show that methods based on cross-distribution entropy may be efficient to stress the decrease in navigation capacities with age. The results are very encouraging for our future goal of adding medical benefits to a leisure equipment. They also show the value of using virtual reality to study the impact of age.
Keywords: Multiscale cross-entropy; Age; Time series; Irregularity; Complexity
Lien vers l'articleJamin, A.; Humeau-Heurtier, A. "(Multiscale) Cross-Entropy Methods: A Review". Entropy 2020, 22, 45.
Abstract: Cross-entropy was introduced in 1996 to quantify the degree of asynchronism between two time series. In 2009, a multiscale cross-entropy measure was proposed to analyze the dynamical characteristics of the coupling behavior between two sequences on multiple scales. Since their introductions, many improvements and other methods have been developed. In this review we offer a state-of-the-art on cross-entropy measures and their multiscale approaches.
Keywords: cross-entropy; multiscale cross-entropy; asynchrony; complexity; coupling; cross-sample entropy; cross-approximate entropy; cross-distribution entropy; cross-fuzzy entropy; cross-conditional entropy
Lien vers l'articleRésumé :
Contexte. L’activité physique, notamment en double-tâche, est préconisée pourprévenir les chutes des personnes âgées mais son manque d’attrait limite son adhésion.
Objectif de la recherche. Créer un programme de prévention attrayant, sur 8 semaines, via des "balades à vélo" en réalité virtuelle, sur le vélo CycléoONE.
Méthodologie. Nos critères étaient : 1) des séances bihebdomadaires comportant 3 balades de 10 min chacune, dans des environnements variés, stimulant respectivement les capacités physiques, cognitives et les interactions sociales, et incluant diverses double-tâches ; 2) des balades écologiques avec augmentation progressive des différentes capacités. Deux résidants d’un EHPAD ont participé à la création de ce programme : bilans sur le vélo, suivi de 5 séances d’entraînement en binôme.
Résultats. Les capacités physiques ont été stimulées en variant notamment la puissance du pédalage, et les interactions sociales via des échanges lors des balades. Pour les balades cognitives, diverses situations de vie quotidienne ont été spécifiquement créées. L’activité vélo en réalité virtuelle a été très appréciée. Les deux participants ont déjà amélioré différentes capacités visées au bout de 5 séances.
Discussion/conclusion. Au vu de ces données prometteuses, l’efficacité de ce programme pour prévenir les chutes des personnes âgées sera prochainement testée.
Mots-clés : Personne âgée ; activité physique ; cognition ; environnement virtuel ; prévention des chutes ; participation sociale
Abstract: Aging is a complex phenomenon that can be studied by analyzing age-related alterations in the execution of various tasks. This work deals with the processing of navigation data acquired from a bike simulator in two populations: young healthy subjects and older adults with loss of autonomy. Our goal is to analyze the influence of age in the ability to ride a bike on a virtual straight line with the perspective to be able to identify diseases in a geriatric population. For this purpose we process time series defined as the deviation from the center of the virtual path, in two straight lines: for each time series we quantify the irregularity – through sample entropy (SampEn) – of intrinsic mode functions obtained from complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). The results show that our approach is able to reveal different learning and adaptation skills in the two populations. Our work could be useful to help the geriatricians in their diagnoses.
Keywords: Empirical mode decomposition; CEEMDAN; irregularity; sample entropy; bike simulator; age
Abstract: We present herein a new approach - the so-called time-shift multiscale cross-distribution entropy (TSMCDE) - to quantify the complexity between two sequences. By analyzing biomedical data, we reveal that TSMCDE over-performs other cross-entropy measures. Thus, TSMCDE, multiscale cross-sample entropy (MCSE), multiscale cross-distribution entropy (MCDE), and time-shift multiscale cross-sample entropy (TSMCSE) were applied to handlebar angle and speed time series recorded from a bike simulator. Twenty-four subjects divided into two groups (12 subjects each) participated to the study. The first group corresponds to young healthy subjects. The second group corresponds to older adults with loss of autonomy. Our results show that a link may exist between complexity and the age and physical state of a population. Moreover, TSMCDE leads to a better differentiation of the two groups than MCSE, MCDE, and TSMCSE. TSMCDE should now be tested on other types of data and on larger datasets to prove its usefulness and its efficiency.
Abstract: This study deals with the telemonitoring, with a connected tensiometer, of 16 patients treated for a kidney cancer. Each one of these patients recorded his/her blood pressure at home during 63 days and the data was sent to his/her medical doctor. At the same time they were treated with antihypertensive medication when necessary. In this work, our goal was to analyze the complexity of the blood pressure time series. For this purpose, we proposed to use the refined composite multiscale entropy (RCMSE) measures. Our results show that the patterns of RCMSE through temporal scales evolve with the antihypertensive medication. The later might therefore have an impact on home-acquired blood pressure complexity. RCMSE could therefore be an interesting information theory-based tool to study home-acquired physiological data.
Keywords: telemonitoring; connected tensiometer; blood pressure; time series; multiscale entropy; clustering; irregularity; complexity
Abstract: In this paper, we detail an in-home aggregation plateform for monitoring physiological parameters, and involving two objective physical sensors (bio-impedanceter and thermometer) and a subjective one (fatigue level perceived by the patient). This plateform uses modern IoT-related technologies such as embedded systems (Raspberry Pi and Arduino) and the MQTT communication protocol. Compared to many related works, monitoring is enterely achieved using a box as a central element, while the mobile device (tablet) is only used for controlling the acquisition procedure using a simple web browser, without any specific application. An example of a time stamped set of acquired data is shown, based on the in-home monitoring of healthy volunteers.
Keywords: IoT; healthcare; bio-impedancemetry; MQTT; raspberry Pi; arduino