Yuan-Ting Zhang · Paulo Carvalho Ratko Magjarevic (Eds.)
Volume 64 International Conference on Biomedical and Health Informatics ICBHI 2015, Haikou, China, 8–10 October 2015
IFMBE Proceedings Volume 64 Series editor Ratko Magjarevic
Deputy Editors Fatimah Ibrahim Igor Lacković Piotr Ładyżyński Emilio Sacristan Rock
The IFMBE Proceedings book series presents the results of IFMBE Conferences. These scientific conferences deal with various topics of medical, biological and clinical engineering, and biophysics. They are organized or endorsed by the International Federation for Medical and Biological Engineering (IFMBE). The aims of the IFMBE conferences are to encourage research and the application of knowledge, and to disseminate information and promote collaboration. The papers of the IFMBE proceedings present research results of a high impact for the community and their high scientific standard is guaranteed by a double peer-reviewing of every published paper. The topics included but are not limited to: • • • • • • • •
Diagnostic Imaging, Image Processing, Biosignal Processing Modeling and Simulation, Biomechanics Biomaterials, Cellular and Tissue Engineering Information and Communication in Medicine, Telemedicine and e-Health Instrumentation and Clinical Engineering Surgery, Minimal Invasive Interventions, Endoscopy and Image Guided Therapy Audiology, Ophthalmology, Emergency and Dental Medicine Applications Radiology, Radiation Oncology and Biological Effects of Radiation
The IFMBE Proceedings series is an official publication of the International Federation for Medical and Biological Engineering. IFMBE Proceedings are indexed by Google scholar. Thomson Reuters and Scopus index many volumes in their ISI Proceedings and the Scopus database, respectively.
More information about this series at http://www.springer.com/series/7403
Yuan-Ting Zhang • Paulo Carvalho Ratko Magjarevic Editors
International Conference on Biomedical and Health Informatics ICBHI 2015, Haikou, China, 8–10 October 2015
123
Editors Yuan-Ting Zhang The Chinese University of Hong Kong Hong Kong, China
Ratko Magjarevic Faculty of Electrical Engineering and Computing University of Zagreb Zagreb, Croatia
Paulo Carvalho Universidade De Coimbra Coimbra, Portugal
ISSN 1680-0737 ISSN 1433-9277 (electronic) IFMBE Proceedings ISBN 978-981-10-4504-2 ISBN 978-981-10-4505-9 (eBook) https://doi.org/10.1007/978-981-10-4505-9 Library of Congress Control Number: 2018954030 © Springer Nature Singapore Pte Ltd. 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore
Contents
Inter-limb Coordination Assessment and Fall Risk in ADL . . . . . . . . . . . . . . . . . . . Tomislav Pozaic, Anna-Karina Grebe, Michael Grollmuss, Nino Haeberlen, and Wilhelm Stork Optimization of the Amplicons Detection System of Loop-Mediated Isothermal Amplification on Microfluidic Compact Disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shah Mukim Uddin, Fatimah Ibrahim, Jongman Cho, and Kwai Lin Thong Implementation of an Electronic Prescription System for Ambulatory Care . . . . . Marc Nyssen, and Yelina Piedra Quantitative Coronary Analysis Using 3D Coronary Reconstruction Based on Two Biplane Angiographic Images: A Validation Study . . . . . . . . . . . . . . . . . . . Panagiotis K. Siogkas, Lambros S. Athanasiou, Antonis I. Sakellarios, Kostas A. Stefanou, Themis P. Exarchos, Michail I. Papafaklis, Katerina K. Naka, Lampros K. Michalis, and Dimitrios I. Fotiadis mHealth Platform for Parkinson’s Disease Management . . . . . . . . . . . . . . . . . . . . . Dimitrios Gatsios, George Rigas, Dragana Miljkovic, Barbara Koroušić Seljak, Marko Bohanec, Maria T. Arredondo, Angelo Antonini, Spyros Konitsiotis, and Dimitrios I. Fotiadis
1
7 13
21
29
Design of a Serious Game to Increase Physical Activity by Adding Direct Benefits to the Game for Conducting Sport Activities . . . . . . . . . . . . . . . . . . . . . . . René Baranyi, Dennis M. Binder, Nadja Lederer, and Thomas Grechenig
37
An Adaptive Compression Algorithm for Wireless Sensor Network Based on Piecewise Linear Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jia-Heng Li, Xiao-Lin Zhou, Rong-Chao Peng, and Feng Lv
43
Improving the Accuracy of the KNN Method When Using an Even Number K of Neighbors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alberto Palacios Pawlovsky and Daisuke Kurematsu
49
Effectiveness of Evidence-Based Venous Thromboembolism Electronic Order Sets Measured by Health Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jacob Krive, Joel S. Shoolin, and Steven D. Zink
57
The Diabino System: Temporal Pattern Mining from Diabetes Healthcare and Daily Self-monitoring Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Eleni I. Georga, Vasilios C. Protopappas, Eleni Arvaniti, and Dimitrios I. Fotiadis
61
Adaptive Latent Space Domain Transfer for Atrial Fibrillation Detection . . . . . . . Xing-Bin Qin, Yan Yan, Jianping Fan, and Lei Wang Detection of Chewing Motion Using a Glasses Mounted Accelerometer Towards Monitoring of Food Intake Events in the Elderly . . . . . . . . . . . . . . . . . . . Gert Mertes, Hans Hallez, Tom Croonenborghs, and Bart Vanrumste
67
73
v
vi
Intra-operative Tumor Tracking Using Optical Flow and Fluorescent Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Daniel Y. Kim, John H. Phan, and May D. Wang Measuring Physiological Stress Using Heart-Related Measures . . . . . . . . . . . . . . . . An Luo, Siyi Deng, Michael J. Pesavento, and Joseph N. Mak Associating Protein Interactions with Disease Comorbidity to Prioritize Colorectal Cancer Genes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sayedeh Razieh Abdollahi Demneh, Sama Goliaei, and Zahra Razaghi Moghadam Synchronization Analysis of EEG Using the Hilbert Huang Coherence . . . . . . . . . Eiji Kondo, Masatake Akutagawa, Takahiro Emoto, Yoshio Kaji, Fumio Shichijo, Kazuhiko Furukawa, Hirofumi Nagashino, Shinsuke Konaka, and Yohsuke Kinouchi
Contents
79 85
91 97
Identifying Cancer Subnetwork Markers Using Game Theory Method . . . . . . . . . 105 Saman Farahmand, Sama Goliaei, Zahra Razaghi Moghadam Kashani, and Sina Farahmand On Fabrication of a Shoe Insole: 3D Scanning Using a Smartphone . . . . . . . . . . . 111 Tomislav Pribanić, Tomislav Petković, Matea Đonlić, and Vedran Hrgetić Development of an Electronic Patch for Falls Detection and Elderly Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 Bouchta Hajjine, Christophe Escriba, Eric Campo, Sabeha Fettouma Zedek, Pascal Acco, Georges Soto-Romero, Anne Hemeryck, and Jean-Yves Fourniols Spatial Interactions of Electrically Evoked Potentials in Visual Cortex Induced by Multi-retinal Electrical Stimulation in Rats . . . . . . . . . . . . . . . . . . . . . . 123 Hui Xie, Yi Wang, and Leanne Lai-Hang Chan An Approach for Body Motion Registration Using Flexible Piezoelectret Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Rui Xu, Qifang Zhuo, Xiangxin Li, Haoshi Zhang, Yanhu Cai, Lan Tian, Xiaoqing Zhang, Peng Fang, and Guanglin Li Indocyanine Green Loaded, PEGylated, Reduced Graphene Oxide as a Highly Efficient Passive Targeting Contrast Agent for Photoacoustic/ Fluorescence Dual-Modality Tumor Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Jingqin Chen, Chengbo Liu, and Liang Song Photoacoustic/Ultrasonic Dual-Modality Endoscopy in Vivo . . . . . . . . . . . . . . . . . . 137 Riqiang Lin, Yan Li, Jianhua Chen, and Liang Song A Novel Compact Linear-Array Based Photoacoustic Handheld Probe Towards Clinical Translation for Sentinel Lymph Node Mapping . . . . . . . . . . . . . 139 Mucong Li, Chengbo Liu, and Liang Song Estimating Blood Pressure with a Smartphone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 Rong-Chao Peng, Wen-Rong Yan, and Xiao-Lin Zhou Is Beat-to-Beat Blood Pressure Variability in Frequency Domain Associated with the Occurrence of Carotid Plaques? . . . . . . . . . . . . . . . . . . . . . . . . 143 Dan Wu, Huahua Xiong, Yujie Chen, Heye Zhang, and Yuan-Ting Zhang Power Aware Topology Management and Congestion Control Mechanism in High Medical QoS WHMNs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Fangmin Sun and Ye Li
Contents
vii
DTI Quantitative Analysis on Microstructural Abnormality in Post Stroke Depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Chenfei Ye, Jun Wu, Xuhui Chen, and Heather T. Ma Marrow Fat Effect on Trabecular Biomechanics in Different BV/TV Subjects—A Simulation Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Yang Chen, Liang Li, James F. Griffith, Ping-Chung Leung, and Heather T. Ma A New Atlas Pre-selection Approach for Multi-atlas Based Brain Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Hengtong Li, Chenfei Ye, Jingbo Ma, Susumu Mori, and Heather T. Ma A Study of Alzheimer’s Disease Based on DTI Gaussian Mixture Analysis . . . . . . 155 Jingbo Ma, Chun Sing Wong, and Heather T. Ma SVM-Based Approach for Human Daily Motion Recognition . . . . . . . . . . . . . . . . . 157 Haitao Yang, Xinrong Zhang, Mengting Chen, and Heather T. Ma High-Speed Intravascular Spectroscopic Photoacoustic Imaging at Two Spectral Bands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Xiaojing Gong, Yan Li, Riqiang Lin, Ji Leng, and Liang Song A Hybrid Non-invasive Method for the Classification of Amputee’s Hand and Wrist Movements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 Oluwarotimi Williams Samuel, Xiangxin Li, Xu Zhang, Hui Wang, and Guanglin Li The Research and FPGA Implementation of ECG Signal Preprocessing . . . . . . . . 167 Wenjun Su, Yunping Liang, Mengni Li, and Ye Li Comparison of the Correlation of Different Pulse Transit Time Parameters to Blood Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Wan-Hua Lin, Oluwarotimi Williams Samuel, Qing Liu, Yuan-Ting Zhang, and Guanglin Li Relative Analysis Between Curative Effect Evaluation and Electroencephalograph of Stroke Patient in Convalescence . . . . . . . . . . . . . . . . . . . 175 Xiao-Mao Fan, Xing-Xian Huang, Ye Li, Hai-Bo Yu, and Yun-Peng Cai Big Data Analysis of Hypertension Complications Bases on Shenzhen Medical Information Management Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 Yu-Jie Yang, Qi Li, and Yun-Peng Cai Vital Signs Analysis for Oceanauts in Deep Sea Submerged Environment: A Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 Fen Miao, Ye Li, and Lu Shi Correlation Analysis of the Time Difference Between Multi-wavelength PPG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 Jing Liu and Yuan-Ting Zhang Epidermal Bioelectronics Toward Oximetry and Health Care Applications . . . . . . 183 Jie Zhang, Huihua Xu, Ningqi Luo, and Ni Zhao An Investigation of Time Difference Between Epidermal Pressure Pulse and PPG Signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 Wen-Xuan Dai, Ni Zhao, and Yuan-Ting Zhang Automatic Co-registration of MEG-MRI Data Using Multiple RGB-D Cameras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Shih-Yen Lin, Chin-Han Cheng, Li-Fen Chen, and Yong-Sheng Chen
viii
Comparison of Heart Rate Variability and Pulse Rate Variability of Respiratory Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 Yi Han, Wen-Chen Lin, Sheng-Cheng Huang, Cheng-Lun Tsai, and Kang-Ping Lin Pilot Project: ICT System for Management and Self-management of Diabetes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 Sara Zulj, Luka Celic, Mladen Grgurevic, Manja Prasek, and Ratko Magjarevic Detection of Atrial Fibrillation Using 12-Lead ECG for Mobile Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Ricardo Jorge dos Santos Couceiro, Paulo Carvalho, Jorge Henriques, Rui Paiva, and Manuel Jesus Antunes A Multi-feature Approach for Noise Detection in Lung Sounds . . . . . . . . . . . . . . . 211 Adriana Leal, César Teixeira, Ioanna Chouvarda, Nicos Maglaveras, Jorge Henriques, Rui Paiva, and Paulo Carvalho Reconstruction and in Silico Simulation Towards Electricigens Metabolic Network of Electronic Mediator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Yuhe Wang, Zhenglin Tong, and Jianming Xie An Attempt to Define the Pulse Transit Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 Xiao-Rong Ding, Jing Liu, Wen-Xuan Dai, Paulo Carvalho, Ratko Magjarević, and Yuan-Ting Zhang
Contents
Inter-limb Coordination Assessment and Fall Risk in ADL Tomislav Pozaic, Anna-Karina Grebe, Michael Grollmuss, Nino Haeberlen, and Wilhelm Stork
Abstract
Fall risk assessment research has largely been focused on individual biomechanical measures or assessment in clinical setting. The goal of the study was to evaluate the fall risk from the inertial sensor data from activities of daily living (ADL) based on the inter-limb coordination assessment. Eight older adults with higher risk of falling and eight adults with no risk of falling were monitored for one week with hip and wrist sensor node. A one-way analysis of variance and 95% confidence interval were applied to investigate associations between extracted temporal inter-limb coordination measures for these two groups. Results have shown significantly higher asymmetry in lower limbs and between contralateral arm and leg for subjects with higher risk of falling, allowing us to reliably distinguish these two groups.
1
Introduction
A leading cause of injuries in older adults is falling, causing a heavy burden on the health care system. Each year, one in every three adults aged over 65 falls [1]. Fall risk factors can be divided into five domains [2]: sociodemographic factors, medical and psychological factors, medication risk factors, mobility factors and sensory risk factors. Epidemiological studies have shown that transitions and walking are main mobility fall risk factors causing 41% and 36% of all falls respectively [3]. Previous studies have investigated mobility factors in terms of variability of biomechanical measures, such as
stride time, walking speed, stride length, stance and swing times and individual joint kinematics [4, 5]. The timing of gait events in the lower limbs is more asymmetric and less stable in older adults [6]. Additionally, gait stability and inter-limb coordination are very well correlated and as such good indicators for falls [7]. Inter-limb coordination primarily involves movements requiring sequential and simultaneous use of both sides of the body with a high degree of rhythmicity. More precisely, it involves the timing of motor cycles of the limbs in relation to one another [8]. Such actions are commonly divided into two categories [9]: bimanual coordination (involves skilled inter-limb coordination of the two arms or legs in a bimanual action) and coordination between hands and feet (involves the simultaneous coupling of the upper and lower limbs). Many previous studies focused only on lower extremities [10], assessed inter-limb coordination using camera based tools [11] or in clinical settings during short periods of time [12, 13]. While camera assessment tools are expensive and subject of privacy concerns, clinical assessments oversimplifies geriatric fall risk, which can be more accurately described with fuzzy boundaries as a multifactorial disorder. Our work focuses on assessment of coordination of upper and lower extremities in terms of fall risk in home environment during ADL using a system of two sensor nodes. A novel approach for inter-limb coordination assessment enables optimization of the number of sensors, thus enabling unobtrusive, user-friendly measurement during a longer period of time (one week). We hypothesize that fall-prone adults will have less coordinated movements and higher variability of gait than adults with no risk of falling on a weekly basis. Using features that uniquely describe interlimb coordination it should be possible to reliably distinguish between these two groups.
T. Pozaic (&) A.-K. Grebe M. Grollmuss N. Haeberlen Bosch Healthcare Solutions GmbH, Waiblingen, Germany e-mail:
[email protected] W. Stork Karlsruhe Institute of Technology, Karlsruhe, Germany © Springer Nature Singapore Pte Ltd. 2019 Y.-T. Zhang et al. (eds.), International Conference on Biomedical and Health Informatics, IFMBE Proceedings 64, https://doi.org/10.1007/978-981-10-4505-9_1
1
2
T. Pozaic et al.
Table 1 Subject characteristics Characteristic
Fallers
Non-fallers
N
8
8
Female (%)
75
25
Age (years)
68.5 ± 8.8
57.2 ± 6.5
Height (cm)
166.0 ± 13.8
173.9 ± 8.3
Weight (kg)
75.6 ± 18.4
81 ± 12.5
BMI, (kg/m )
27.7 ± 7.7
26.8 ± 3.9
SOMC (0–28)b
2.0 ± 2.4
1.5 ± 3.5
2 a
Habitual gait speed, (m/s)
1.0 ± 0.2
1.1 ± 0.2
History of falls
0.8 ± 0.7
0
FRAQc
7.4 ± 2.6
1.4 ± 1.2
a
BMI Body Mass Index b SOMC Short Orientation-Memory-Concentration test c FRAQ Fall Risk Assessment Questionnaire
2
Methods
2.1 Subjects Sixteen older adults were recruited with the characteristics as shown in Table 1. These show the mean values with corresponding standard deviations. The study was approved by the Ethical Committee of the Medical Faculty and the University Hospital of Tübingen. All participants signed informed consent according to the Declaration of Helsinki. Exclusion criteria for the study were more than 10 points on Short Orientation-Memory-Concentration (SOMC) test (cognitive impaired subjects), inability to walk or terminal diseases.
2.2 Experiment Setup Subjects were wearing two sensor nodes, one attached on the wrist and one on the ipsilateral hip. Each sensor node consisted of a 3-axis accelerometer BMA280, gyroscope BMG160 and magnetometer BMC055 (all produced by Bosch Sensortec GmbH). Accelerometer measurement range was set to ±4 g, gyroscope to ±500°/s and magnetometer to ±1000 µT. 14-bit accelerometer and 16-bit gyroscope and magnetometer data were sampled with a sampling frequency of 100 Hz. Each data package was transmitted over a Bluetooth Low Energy (BLE) connection to an Android phone (LG2 mini) attached at the belt around the waist. Phone was requesting the time from the sensor nodes every minute and correlated it with the real time. The round trip between the phone and sensors is integrated into this mapping for synchronization purposes. Data corresponding to one measurement day was stored in one file in order to enable easier offline processing.
Data acquisition was performed for one week (seven consecutive days) in the home environment of each subject. Rechargeable lithium battery (170 mAh) supplying the sensor node lasts for approximately 8 h. Subjects put on the sensors in the morning and wore them during activities of daily living. The battery and phone are charged over night. On the first day of the measurement week the subject’s supervisors collected anthropometric measures. The habitual gait speed was determined by letting the subjects walk on a straight line not shorter than 3.5 m and measuring the elapsed time. The length of the walking line was adjusted to various conditions in subject’s home (e.g. small apartments, obstacles etc.). Additionally subjects answered a fall risk assessment questionnaire (FRAQ) containing 18 most significant factors for risk of falling indentified in [2]. Answers were graded with either 0 or 1 (depending if the indentified risk factor was present or not), except the number of prescript medications and number of falls in the last 12 months. Total score defined as the sum of all answers was used to split the subjects into two groups: fallers and non-fallers. Fallers are subjects with a total score of four or higher, while non-fallers are subjects with a total score lower than four points.
3
Data Analysis
3.1 Preprocessing Data was preprocessed offline using MATLAB R2012b. Sensor data for the wrist and hip were synchronized using a (proprietary) method for mapping the phone and sensor times. Signals, particularly from the wrist, were affected with data loss due to various artifacts in BLE connection (e.g. various obstacles, distance between sensors and phone). Thus, before further processing, the missing data was interpolated using linear interpolation. Missing data is detected when the time difference between two consecutive BLE packages ti and ti1 is ti ti1 [ 1:5 * TS , where TS is defined sampling period. This action maintained data from both sensors synchronized during the whole day and enabled sample wise signal processing. Inertial sensors (accelerometer, gyroscope and magnetometer) provide data in three perpendicular axes of their local coordination system. To describe the orientation of sensor’s local system in relation to the geodetic coordination system approach with Euler angles (yaw, pitch and roll) estimated from all three inertial sensors was used. The relation is described with a Yaw-Pitch-Roll angle rotation matrix MYPR defined by standard convention (“x-convention”) rotation order:
Inter-limb Coordination Assessment and Fall Risk in ADL 1 0 at ¼ MYPR at ;
3
ð1Þ
where a0t is one sensor sample at moment t and at is the corresponding value in the geodetic coordination system. Solving (1) for the acceleration signals the linear acceleration of human movement is calculated.
3.2 Walking Bouts Acceleration based step detector was previously developed in C and used as binary mex file for processing sensor data in MATLAB. The step detector algorithm is based on an adaptive threshold approach over a sliding window of 300 samples. A walking bout is defined as the time between start and end of walking. Start of walking is depicted with three or more consecutively detected steps, while end of walking is determined when no steps for maximum step duration time (3 s) were detected. Data for each day for each subject was processed and only walking bouts longer than 10 s were taken into consideration. Despite the fact that high step rate variability is present for short walking bouts, they were still taken into consideration due to the older age of subjects and the relatively small number of long walking bouts (100 or more steps).
3.3 Gait Speed Linear acceleration (defined as an acceleration in the direction of human movement), was used to calculate gait speed of a particular walking bout. In order to remove tilt from the acceleration signal, a following filter proposed in [14] was applied: xoutput ¼
xa xa ; cosða sinðxa ÞÞ
ð2Þ
where xa is linear acceleration signal and xoutput is signal without tilt. To reduce additionally the signal drift due to high frequency noise, data was filtered using moving average filter with window size of one second. The gait velocity was then calculated by integrating the filtered acceleration signal for each walking bout and then averaged over the particular interval. Previous systematic review of gait speed values for long term care residents, which are at higher risk of falling, has shown that usual gait speed in clinical setting is 0.58 m/s [15]. Moreover, slow gait speeds have also been related to
higher risk of institutionalization and mortality. Thus, only walking bouts with mean gait speed lower than 0.6 m/s were taken into further consideration.
3.4 Inter-limb Coordination Features Features extracted in this study for assessment of inter-limb coordination and are as follows: • • • • •
inter-limb coordination index (IC), ipsilateral coordination index (YC), contralateral coordination index (CC), step time variability (STV), swing phase time variability (SPV).
STV and SPV features describe bimanual coordination, while IC, YC and CC features describe hands/feet coordination. Gait cycle of human walking can be split into swing and stance phase, where swing phase starts with toe off event and ends with heel strike. Toe off event was detected from the hip acceleration signal as the first local minima following the heel strike of the opposite foot. SPV is defined as variability of swing time duration of steps in each walking bout. STV is defined as a variability of duration of steps detected for each walking bout. IC parameter is defined for each walking bout as a mean time delay between highest backward or forward point in arm swing and corresponding heel strike. Highest backward point in the arm swing fits to the heel strike of ipsilateral foot, while highest forward point in arm swing fits to heel strike of contralateral foot. Since it was defined that sensor should be worn always on the same hip (subject arbitrarily chooses on first day of measurement on which side the sensors will be worn), it is possible to distinguish between left and right heel strikes based on the gyroscope signal. Namely, empirically it has been noticed that rotation of the hip while walking happens before the heel strike and it is visible in the gyroscope signal (angular velocity) on the side where sensor is worn. By calculating the area under curve of the gyroscope signal between two consecutive heel strikes, left and right steps were distinguished. YC and CC features were then calculated as the mean time delay between arm swing and ipsilateral and contralateral heel strike, respectively.
3.5 Statistical Analysis Analysis of the features was performed on a weekly basis, meaning that feature values for each particular subject were averaged for all walking bouts satisfying above described
4
T. Pozaic et al.
conditions over the whole week (or at least for all days when recording was successfully performed). A one-way analysis of variance (ANOVA) was used to analyze the difference between two defined groups for all five features (significance, p). The mean (µ) values were extracted for each feature together with corresponding 95% confidence interval (95% CI). MATLAB built-in functions anova1 and paramci were used for statistical analysis.
0,43
p