Integrating Multi-Sensors for Observing Post ACL Reconstruction Recovery Progress PDF

Preview

Summary

TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 4, PP. 171-181, APR., 2013. Integrating Multi-Sensors for Observing Post ACL Reconstruction Recovery Progress SMN Arosha Senanayake 1, Owais Ahmed Malik 2, and Pg. Mohammad Iskandar 3,  Abstract: This study investigates the integration of w...

Description

Accelerat ing t he world's research.

Integrating Multi-Sensors for Observing Post ACL Reconstruction Recovery Progress Owais Malik Transaction on Control and Mechanical Systems

Cite this paper

Downloaded from Academia.edu 

Get the citation in MLA, APA, or Chicago styles

Related papers

Download a PDF Pack of t he best relat ed papers 

Assessing post -ant erior cruciat e ligament reconst ruct ion ambulat ion using wireless wearabl… SM Namal Arosha Senanayake, Owais Malik A Mult isensor Int egrat ion-Based Complement ary Tool for Monit oring Recovery Progress of Ant erior C… SM Namal Arosha Senanayake Wireless Mult i-Sensor Int egrat ion for ACL Rehabilit at ion Using Biofeedback Mechanism Owais A Malik

TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 4, PP. 171-181, APR., 2013.

Integrating Multi-Sensors for Observing Post ACL Reconstruction Recovery Progress

SMN Arosha Senanayake 1, Owais Ahmed Malik 2, and Pg. Mohammad Iskandar 3,  Abstract: This study investigates the integration of wireless micro-electro-mechanical-systems (MEMS) and electromyography (EMG) sensors for developing a motion analysis system for post-operative recovery monitoring of Anterior Cruciate Ligament (ACL) reconstructed subjects based on biofeedback mechanism. The kinematics and neuromuscular signals have been combined using mixed signal processing techniques and a feature set has been generated for classification of recovery status of subjects. Two intelligent techniques (Feed-forward Artificial Neural Network and Fuzzy Rule-based Classifier) have been tested and compared for providing rehabilitation status of the subjects. The system has been tested on a group of national athletes and it provides an un-obstructive assessment of the kinematics and neuromuscular changes occurring after ACL reconstruction in an athlete. The successful implementation and testing of multimodal sensors' integration show its feasibility in identifying the clinical stage of the recovery process of athletes after ACL repair and using it as an assistive tool for clinical decision-making during the rehabilitation regimen. Keywords: Wireless Sensors, Signal Processing, Multimodal Integration, Anterior Cruciate Ligament, Rehabilitation. 1

1. INTRODUCTION With the advent of small scale, light weight and wearable sensors, different types of bio-signals are readily available for monitoring health and performance related activities of subjects/athletes [1-4]. The availability of low-cost solutions in terms of light weight miniaturized sensors has given rise to pervasive healthcare systems. These miniature sensors provide quantitative information that can be utilized to objectively monitor the fitness condition and functional recovery status of sportspersons in both clinical and home settings at real time as well as offline [5] . The usage of different sensors has been investigated in recent studies for observing the changes in knee joint dynamics and gait pattern during rehabilitation process after knee surgery [6-8]. These sensors measure kinematics,

kinetics and/or neuromuscular behavior of knee by employing different mechanisms. MEMS inertial sensors, consisting of single or multiple accelerometers and gyroscopes, have been used for activity recognition in different domains including medical, sports and home monitoring and assisted living applications. An ear worn activity recognition (e-AR) sensor based on accelerometer has proved to be quite successful in observing recovery after knee replacement surgery [7]. Different post-operative activities can be continuously monitored using such miniaturized sensor which later on may help in quantifying any impairment after knee arthroplasty. A motion capture system based on arrays of magnetic, angular rate and gravitational (MARG) sensor has also been used to measure the joint angles for identifying the gait patterns for rehabilitation after total knee replacement surgery [9-10]. The measurement done by these inertial sensors are reasonably comparable with high quality but expensive motion capture systems (e.g. OPTOTRAK and Vicon) [11]. In addition to MEMS sensors, the role of electromyography (EMG) has been demonstrated for quite long in monitoring the neuromuscular activity in knee rehabilitation (ACL-injured knee) [12-13]. The studies show that the muscle recruitment and response to Anterior Tibial Translation (ATT) vary significantly in ACL deficient and normal subjects, which directly affect the ACL deficient individual‟s physical activity and functional performance [14] . In order to monitor such neuromuscular activity, surface electromyography (sEMG) has been used that records the muscle functions non-invasively. In one of the recent studies, Boucher et al [15] have shown that the knee extension active range of motion (AROM) function can be improved by using sEMG triggered neuromuscular electric stimulation (NMES) during post-operative knee rehabilitative exercises. In [16], based upon four studies of level 2 evidence or higher, it has been shown that there is an inconsistency to support the use of EMG biofeedback in knee rehabilitation when used in conjunction with conventional exercise compared to conventional exercise alone.

This research is supported by the University Research Council (URC) grant scheme at the University of Brunei under the grant No: UBD/PNC2/2/RG/1(195) with the title “Integrated Motion Analysis System (IMAS)”. 1 SMN Arosha Senanayake, Universiti Brunei Darussalam, Faculty of Science, Jalan Tungku Link, Gadong BE1410, Brunei Darussalam ([email protected]) 2,3 Owais Ahmed Malik and Pg. Mohammad Iskandar (),Universiti Brunei Darussalam, Faculty of Science, Jalan Tungku Link, Gadong BE1410, Brunei Darussalam. ([email protected], [email protected] ) RECEIVED: 22, FEB., 2013; REVISED: 14, APR., 2013; ACCEPTED: 15, APR., 2013; PUBLISHED: 18, APR., 2013.

ISSN: 2345-234X

172

TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 4, PP. 171-181, APR., 2013.

Nevertheless, sEMG biofeedback intervention still may be used as a tool to involve the patient in the rehabilitation process and to demonstrate the activity within the quadriceps muscle group as it is a non-invasive procedure and has little to no risk. Other than knee rehabilitation, EMG has also been used for identifying the ACL ruptures beforehand for preventing such injuries. It has been shown that athletes with low knee flexor EMG pre-activity and high knee extensor EMG pre-activity during side-cutting are more prone to ACL rupture [17]. Additionally, fabric sensors have also been used in monitoring sports performance of athletes and rehabilitation of knee injuries [1, 18]. However, these individual sensors have their own capabilities and limitations in providing enough information about various human physiologic processes and body activities. For example the inertial sensors can provide kinematics of an athlete but no information about the strength of muscles and their activation timings can be detected by using these sensors. On the other hand, the neuromuscular activities can be monitored by using electromyography (EMG) but it does not provide any information about the joint moments and forces exerted by different body parts which can be recorded by using force plates or other similar sensors. A more holistic picture about the athletes‟ rehabilitation process can be obtained by combining bio-signals recorded from different types of sensors (EMG, ECG, EEG signals from BioCapture Systems, joints dynamics and lower extremity activities using micro-scale motion sensors and kinetic signals using force platforms) rather than relying only on limited bio-signals. The sensor integration and/or fusion approach may provide a more detailed analysis of different activities performed by athletes. The sensor integration is commonly used to complete a task by gathering data from multimodal

sensor. In sensor fusion, data recorded from disparate sensory sources is combined into a single representation to acquire more accurate, reliable and complete information about a given situation. Both sensor integration and fusion require to deal with number of issues e.g. processing and interpreting multiple signals which may be homogeneous or heterogeneous in nature, signals having different sampling rates and/or electromechanical delay etc [19]. The integration of different sensors and information fusion has gained attention recently in different areas including monitoring activity in stroke patients [20-21], tremor sensing/suppression [22] and hand gesture recognition [23]. The major focus of these studies has been dealing with movements in upper extremities. This study proposes a recovery monitoring system based on integration of kinematics and EMG signals for knee injury (in particular after ACL reconstruction) for athletes. The aim of this study is to investigate for an objective rehabilitation monitoring system for ACL injury which can assist the clinicians and athletes in examining and observing the recovery progression. This system will also provide biofeedback using intelligent techniques to improve the overall rehabilitation process.

2. METHODOLOGY A. General System Architecture The proposed Hardware/Software co-design architecture for observing recovery progress of ACL reconstructed subjects is illustrated in Fig. 1. The functionality of each component related to this study is described as follows. 1) Hardware Components: The system hardware includes the following major components:

Fig. 1. Framework for Anterior Cruciate Ligament Recovery Analysis Using Integrated Sensors. TRANSACTION SERIES ON ENGINEERING SCIENCES AND TECHNOLOGIES (TSEST) ©

SENANAYAKE et al.: INTEGRATING MULTI-SENSORS FOR OBSERVING POST ACL RECONSTRUCTION RECOVERY PROGRESS.

a) Wireless Micro-electro-mechanical Systems (MEMS) Sensors: These body-mounted miniature sensors are used to measure and record the athlete lower extremity motion, particularly knee activities during motion, in terms of angular rate and linear acceleration. The small size and light-weight sensors do not provide any obstruction in the human motion. b) Wireless EMG Capturing System: The electromyography signals from knee flexors and extensors are captured using BioCapture system for monitoring the neuromuscular activity during rehabilitation process. Additionally, electroencephalography signals may also be monitored to study the role/activity of brain during the motor control. c) Video Camera: The video camera provides visual monitoring of the rehabilitation process. The video signals are synchronized with other sensors to analyze the athletes' lower extremity movements and to estimate different recovery phases during experiments. 2) Software Components: The system software includes the following major components: a) Signal Acquisition Layer (SAL): This layer acquires the signals in raw form from different sensors (motion and EMG sensors) and stores them in the database for processing by next layer. b) Signal Pre-Processing Layer (SPL): This layer processes the raw signals into required format for further processing by the system (e.g. signal filtering, knee angle calculation, signal rectification etc.) c) Signal Synchronization Layer (SSL): This layer synchronizes the signals from different sensors in order to analyze the overall performance about the knee joints and muscles activities during rehabilitation. d) Feature Extraction Layer (FEL): This layer is used to extract features from the knee kinematics and neuromuscular data. These features are used by the intelligent data analyzer layer for classifying the recovery status of ACL reconstructed subjects e) Intelligent Data Analyzer (IDA): After transforming the bio-signals into required format and feature extraction, intelligent algorithms are used for motion pattern classification and prediction during rehabilitation. f) Training Model: A training model assists the IDA in classification and prediction during rehabilitation.

173

g) Recovery Monitoring/Biofeedback Interface: A biofeedback system provides monitoring of the recovery progress based on the intelligent algorithms to clinicians and athletes as well. h) Knowledge Base (KB): The knowledge base is used to store the data/signals at different stages and in different formats for processing and classification. It will also contain the information from training model which will be updated based on new results produced by the intelligent data analyzer. This will help the system in learning from previous experiences and reacting appropriately in the new situations. This study focuses on implementation of the signal acquisition, processing, synchronization, feature extraction and motion classification part of the proposed architecture (Fig. 2). B. Participants In order to analyze data, both healthy and ACL-reconstructed subjects were included in this study. There were 5 healthy male and 5 (3 females and 2 males) unilateral ACL reconstructed subjects recruited for monitoring kinematics and muscular activities during the gait cycle. The healthy subjects were having a mean age of 23.5 years, mean height 163 cm, and mean weight 65 kg. For ACL reconstructed male subjects, the mean age, mean height and mean weight were 25 years, 80 Kg and 178 cm respectively. For ACL reconstructed female subjects, the mean age, mean height and mean weight were 28.5 years, 61 Kg and 164.6 cm respectively. The participants were recruited from University of Brunei, Ministry of Defense and Ministry of Sports in Brunei. Ethical procedures were carried out according to the guidelines approved by University of Brunei Graduate Research Office and Ethics Committee. C. Experimental Setup The experimental data was collected using two sensing units namely KinetiSense (ClevMed, Inc.) and BioCapture (ClevMed, Inc.). The KinetiSense is a bio-kinetic analysis system consisting of a command module, wireless transmission radio and sensor units. Each sensor unit (size: 2.2cm x 1.5cm x 1.25cm) contains a tri-axial MEMS accelerometer and a tri-axial MEMS gyroscope to measure

Fig. 2. Processing and Synchronization of Kinematics and EMG Signals. TRANSACTION SERIES ON ENGINEERING SCIENCES AND TECHNOLOGIES (TSEST) ©

174

TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 4, PP. 171-181, APR., 2013.

3-D linear accelerations and 3-D angular velocities respectively. The data from sensors is wirelessly transferred through USB receiver to the computer where the KinetiSense software records the readings for each experiment. The sampling rate of motion sensors data was 128Hz. Each subject was setup with four motion sensors attached to the his/her right thigh, right shank, left thigh and left shank using flexible bulk and Velcro straps to note the angular rate and accelerations of lower limb extremities. This data was then exported to MATLAB to compute the knee angle and for other processing. The EMG signals were recorded using BioCapture physiological monitoring system consisting of BioRadio and USB receiver. The BioRadio records the EMG signals, does initial processing and then wirelessly transmits them to the computer using USB receiver. For our study, the sampling rate to collect EMG signals was set to 960Hz at 12/16 bit A/D conversion. In order to record surface EMG signals, foam snap electrodes were placed on five different knee extensor and flexor muscles including vastus medialis, vastus lateralis, semitendinosus and gastrocnemius medialis. SENIAM EMG guidelines were followed for skin preparation and electrodes placement [24]. The data recorded by BioCapture was exported to MATLAB for filtering and EMG rectification. D. Data Collection Two sets of experiments were conducted in this study. In the first set, healthy subjects were requested to walk at 2, 3 4 and 5 km/hour for 25-30 seconds on treadmill. In the second set, subjects with ACL-reconstruction were requested to walk on a treadmill at a speed of 2, 3, 4 and 5 km/hour. Both kinematics and EMG signals were recorded at the same time during all experiments. E. Knee Angle Computation and Signals Synchronization The knee flexion/extension measurements were obtained from each motion sensor unit placed on the thigh and shank segments of both legs. The sensors were aligned to provide knee angle about the sagittal plane using angular rate about Z-axis (Fig. 3). The angular rate measurements obtained from the motion sensors (MEMS gyroscopes) were low pass filtered using 6th order Butterworth filter with 3 Hz cut-off frequency before computing the orientations. With respect to the placement of each motion sensor, measurements for zero-referencing were obtained prior to starting the experiment (actual motion) when the subjects were in upright position. These measurements were then subtracted from each angular rate during the experiment. Trapezoidal integration method was applied on angular rates (ωR , ωR ank , ωL and ωL ank of both lower limbs to estimate the orientation ( θR , θR ank , θL and θL ank ) of lower extremity. The estimated orientation of thigh and shank is computed using (1), where θ is the estimated orientation at time t, ω is the angular rate of either thigh or shank at time t and ∆t is the sampling time.

Right

Left

Y Bio-Capture X Motion Sensors X Z

Z ωz

ωz X

Y

Y

Motion Sensors X Z

Z ωz

ωz X

Y

Y

Fig. 3. Motion Sensors‟ Placement on Thighs/Shanks and Co-ordinate System.

∆

(1)

The gyroscope integration drift was corrected using complementary filter by fusing orientation measurements from gyroscope and accelerometer available in motion sensor. The corrected angular measurements were used to compute the knee angle. The EMG signals were recorded using surface EMG electrodes placed at four muscles (knee flexors/extensors: vastus medialis, vastus lateralis, semitendinosus and gastrocnemius medialis) on each leg. The raw EMG data from BioCapture was high and low pass filtered using 4th order Butterworth filter for generating EMG envelops for each muscle. The synchronization of both systems was done using re-sampling and gait cycle detection. An up-sampling was done for motion data from KinetiSense to match with the sampling rate of EMG recording from BioCapture. For motion data, the heel strike (HS) was detected by using shank sagittal angular velocity [25]. The HS was identified by examining the timing character tics of the angular velocity and determining the two minima on either side of a peak in velocity curve where every second minimum indicates the HS event. The anteroposterior acceleration from a 2-D accelerometer, available in BioCapture, was used to identify heel strike event in EMG data to mark the gait cycle and observe the overlapping of knee kinematics and EMG signals of different muscles [26]. F. Motion Classification Using Intelligent Mechanisms For classifying the motion of healthy and post-operated (ACL reconstructed) legs, the multilayer feed-forward artificial neural network (ANN) and fuzzy logic (FL) based classifier have been used in this study [27-28]. The temporal nature of the gait cycle parameters was taken into consideration while implementing FL techniques.

TRANSACTION SERIES ON ENGINEERING SCIENCES AND TECHNOLOGIES (TSEST) ©

Kinematics Signal

SENANAYAKE et al.: INTEGRATING MULTI-SENSORS FOR OBSERVING POST ACL RECONSTRUCTION RECOVERY PROGRESS.

Knee Angle

Neuromuscular Signals

Vastus Lateralis Vastus Medialis Semitendinosus

. . . .

Recovered / Not Recovered

Gastrocnemius Medialis

Fig 4. ANN Architecture for Recovery Status Classification Using Kinematics and Neuromuscular Inputs.

A general architecture for ANN is shown in Fig. 4 for five input parameters and one output parameter. The feed-forward back-propagation par...