Enhancing Vibroarthrography by using Sensor Fusion

Dimitri Kraft

1 a

, Rainer Bader

and Gerald Bieber

3 b

University of Rostock, Rostock, Germany

Unimedizin, Rostock, Germany

Fraunhofer IGD, Rostock, Germany

Keywords:

Endoprosthesis, VAG, Sensors, Mobile Device, Implants, Wear, Accelerometer, Microphone, Vibration.

Abstract:

Natural and artiﬁcial joints of a human body are emitting vibration and sound during the movement. The

sound and vibration pattern of a joint is characteristic and changes due to damage, uneven tread wear, injuries,

or other inﬂuences. Hence, the vibration and sound analysis enables an estimation of the joint condition. This

kind of analysis, vibroarthrography (VAG), allows the analysis of diseases like arthritis or osteoporosis and

might determine trauma, inﬂammation, or misalignment. The classiﬁcation of the vibration and sound data is

very challenging and needs a comprehensive annotated data base. Current existing data bases are very limited

and insufﬁcient for deep learning or artiﬁcial intelligent approaches. In this paper, we describe a new concept

of the design of a vibroarthrography system using a sensor network. We discuss the possible improvements

and we give an outlook for the future work and application ﬁelds of VAG.

1 INTRODUCTION

A joint is the connection between bones in the body,

which link the skeletal system into a functional whole.

They are constructed to allow for different degrees

and types of movement (Kim et al., 2009). Joint dis-

order and diseases have a huge impact on the life of

people and are the main cause for disability of el-

derly people. Disorders and diseases such as arthri-

tis causing huge economic costs for nations. In the

US, the costs for treating arthritis were $303.5 bil-

lion or 1% of the 2013 US Gross Domestic Product

(GDP). One of the most common disease of joints

is osteoarthritis (OA), a degenerative disease where

changes in bones, articular cartilages and soft tissues

occur. OA affecting nearly 10% of the population

worldwide and is frequently observed in elderly peo-

ple (44% in people ¿ 80 years). OA was the second

most costly health condition treated at US hospitals

in 2013. In that year, OA accounted for $16.5 bil-

lion, or 4.3%, of the combined costs for all hospital-

izations (Murphy et al., 2017). Tools and technologies

for quantify the health status of human joints outside

of the clinical setting are investigated by researchers

throughout the past decade include the analysis of vi-

bration caused by human joints, the range of motion

https://orcid.org/0000-0002-0604-5854

https://orcid.org/0000-0003-2496-6232

and gait analysis. The analysis of vibration is called

Vibroarthrography (VAG) (McCoy et al., 1987) or Vi-

bration arthrometry (Kernohan et al., 1990). VAG is

a non-invasive screening tool for vibration and sound

analysis of natural joints which was ﬁrst introduced

by McCoy et al. in 1985 (McCoy et al., 1987) (see

Figure 1). The origin concept is older and referred to

Carl Hueter in 1883 (Hueter, 1883).

2 MOTIVATION

Because of the demographic development, joint dis-

eases are an increasing problem. Within the lifetime,

approx. 90 % of the population will suffer on seri-

ous knee or hip problems, hereby women are more

affected than men. In Germany, the most surgery is

knee joint operation by more than 150,000 times per

year. But often, a knee replacement is not needed and

an artiﬁcial knee joint wound last forever. More than

60 % of joints that had to be replaced are younger

than 7 years. Therefore, an easy to use measurement

system is necessary. The usage of a sensor network

that combines also the capability to measure mobility,

strength, sound, weight, acceleration, and light reﬂec-

tion of the veins enables the assessment of a compre-

hensive representation of the joint and body condi-

tion. A measurement over time provides the determi-

Kraft, D., Bader, R. and Bieber, G.

Enhancing Vibroarthrography by using Sensor Fusion.

DOI: 10.5220/0009098701290135

In Proceedings of the 9th International Conference on Sensor Networks (SENSORNETS 2020), pages 129-135

ISBN: 978-989-758-403-9; ISSN: 2184-4380

129

nation of a trend and allows a prognosis beside the

diagnosis. To achieve this goal, the concept and de-

sign of a sensor system is necessary.

3 RELATED WORK

3.1 Sensor Selection

The analysis of knee joint sound experience a long

history of investigation since 1976. Chu et al. pub-

lished a series of papers investigating the acoustic

pattern of knee joints. The researchers found that

cartilage damage around the knee may be classiﬁed

with acoustic sensors (Chu et al., 1978). Further they

found, that pattern recognition in acoustic signals may

be used to distinguish between normal, rheumatoid

and degenerative knees (Chu et al., 1976). Later in

1985, a research team evaluated the diagnostic po-

tential of vibration arthrography by examining 250

subjects who were undergoing diagnostic arthroscopy

and suggested that VAG will become a signiﬁcant di-

agnostic aid in the clinical evaluation of the locomo-

tive system. In 86% of these cases, a characteristic

signal was obtained. Further, it was possible to iden-

tify the affected side and determine how far posteri-

orly the mensical injury lay (Kernohan et al., 1986).

Figure 1: Subject attached to the recordings apparatus of

VAG (McCoy et al., 1987).

Acceleration sensors show the advantage against mi-

crophones that microphones are sensing ambient

noise from the environment too much and provide

a limited frequency response in the audible range

(Frank et al., 1990). Technologically, the accelera-

tion of the skin is a mechanical wave. Hence, mi-

crophones are sensing airwaves that are generated by

swinging material. The advantage of microphones is

that they are able to detect higher frequencies and that

no direct contact to the body is necessary. In rela-

tion to industrial inspections for mechanical machine

components of wear by vibration analysis (Tandon

and Choudhury, 1999; Orhan et al., 2006), the vibra-

tion and sound analysis of natural and artiﬁcial joints

seems also to be manageable.

Recently, Klemm et al. studied the effects of

sensor placement in their recent study and suggest,

that mechanical extensions to the sensors, like 3D-

printed caps may improve the data acquisition pro-

cess. This ensures a ampliﬁcation and a good trans-

mission of lower frequency sounds (up to 125Hz)

without attenuation of higher bands. During measure-

ment, the sensors, off-the-shelf airborne audio sensors

(SPH0645LM4H-B), were attached to the medial and

lateral side of the knee. The measurement protocol

consists of 20 sit to stand movements cycles (Klemm

et al., 2019).

3.2 Feature Assessments

Several authors carried out research to investigate

acoustic emission to assessing the dynamic integrity

of joints (Prior et al., 2010). To detect acoustic emis-

sion, it is necessary to attach piezoelectric transducers

to the surface of a structure under load. Shark et al.

developed a joint angle based acoustic emission sys-

tem. After the data acquisition process, the authors

applied feature extraction techniques and dimension-

ality reduction to discriminate between healthy and

unhealthy knees (Shark et al., 2010). Shark et

al. found that the OA knees produce consistently

and substantially more acoustic emission events with

higher peak and mean magnitude than healthy knees.

Teague et al. used piezoelectric transducer, elec-

tret amd MEMS microphone, and IMU to assess the

click location of knees during movement. During

their study the researchers evaluated different types

of microphone and carried out a comparison between

on skin ad off skin joint sound signals. Further, they

found that acoustic events occur at consistent joint

angles during repetitive motions of healthy subjects.

They suggest that joint sound measurements are re-

peatable with sensing technology that can be imple-

mented in an inexpensive and wearable form factor

(Teague et al., 2016). In 2018, Andersen et al. ex-

amined the methodical aspects of VAG assessment

by investigating sensor placement and ﬂexion-exten-

sion movements under load. Their study showed, that

VAG parameters are affected by load level, movement

type and the location of the accelerometer over the

knee joint in asymptomatic subjects (Andersen et al.,

2018).

SENSORNETS 2020 - 9th International Conference on Sensor Networks

130

3.3 Classiﬁcation Techniques for VAG

The current research in Vibroarthrography underlie

different problems. Most research in vibroarthrog-

raphy focus on the classiﬁcation process (Rangayyan

et al., 1997; Wu and Krishnan, 2011; Tai et al., 2015;

Nalband et al., 2016), not considering different meth-

ods to acquire the data. It is not clear, which sen-

sors (IMU, MEMS-microphone, piezoelectric trans-

ducer, Ultrasound) should be used to assess the cur-

rent health state. Further, only little research is done

evaluating the optimal sensor placement and envi-

ronmental setting during the measurement. Several

studies showed, that crepitus of the knee is depen-

dent on the angular velocity (Kernohan et al., 1990)

and the load affecting the knee joint during move-

ment(Andersen et al., 2018). The usage of multi-

ple sensors is needed due to the fact, that differ-

ent states of pathology are found in different fre-

quency bands (Frank et al., 1990) and may enable

the early prevention of certain diseases. Peat et al.

suggest that the American College of Rheumatology

(ACR) clinical criteria to classify osteoarthritis seem

to reﬂect later signs in advanced disease. They sus-

pect that other approaches may be needed to iden-

tify early, mild osteoarthritis in the general population

(Peat et al., 2006). The gold standard for diagnosing

certain joint disorder is still X-ray or MRI . Those

techniques are either radiating or expensive, render-

ing the constant monitoring impossible and not use-

ful for early prevention. Abbot et al. concluded that

the intense variability within signals is caused by con-

tacting joint surfaces and forces during motion. This

produce an understandable scepticism in the clinical

community as to the reliability of vibration arthrome-

try and therefore making an adaption in medicine hard

to accomplish (Abbott and Cole, 2013). The perfor-

mance and capabilities of Vibroarthrography in de-

tection of knee disorders are slight behind those of

MRI/MRT. McCauly et al. reported 86% sensitivity

and 74% speciﬁcity in detection of chondromalacia

patellae using MRI (McCauley et al., 1992). Pihlaja-

maki et al. reported 83% sensitivity and 84% speci-

ﬁcity for MRI images for stage III chondromalacia

(Pihlajam

aki et al., 2010).

3.4 Mobile VAG Systems

The need for constant monitoring of health status is

not a new concept, considering the newest technolo-

gies in the Internet of Things area. Several devices

such as smartwatches and ﬁtness tracker measuring

heart rate, respiration rate or general activity of hu-

man beings since a decade. The concept of constant

monitoring of the knee health status, on the other

hand, was ﬁrst introduced a by a research group in

2016. T

oreyin et al. (Toreyin et al., 2016) provided a

prototype and framework to measure the sound made

by joints during daily activities such as sit to stand

transitions, walking or others. Another approach pro-

posed by Msayib et al. in (Msayib et al., 2017) de-

veloped an intelligent monitoring system to measure

the rehabilitation of total knee arthroplasty patients

by assessing the range of motion of the knee. A

recently proposed framework by Athavale and Kris-

han in (Athavale and Krishnan, 2019) is especially

designed for the IoT era. By encoding the signals,

the researchers enabling a efﬁcient way for a con-

stant monitoring of the condition of knee joint with

vibroarthrography (or actigraphy). Unfortunately,

Athavale and Krishan did not propose a effective sen-

sor alignment to measure the knee joint sounds and

vibrations. This work try to outline a concept based

on the aforementioned research.

3.5 Challenges in VAG Systems

Although several researches reported high accuracy,

sensitivity and speciﬁcity (¿95%) in classiﬁcation of

pathological knees via VAG (Kim et al., 2009; Nal-

band et al., 2017; Rangayyan et al., 2013), those re-

sults were obtained on a fairly dated dataset, consist-

ing of 38 abnormal and 51 normal VAG signals ob-

tained from a single axis accelerometer. Scalograms

of abnormal and normal VAG signals are depicted in

Figure 2. Rangayyan et al. suspect in (Rangayyan

et al., 2013) that those remarkable results are obtained

by overﬁtting on a speciﬁc dataset, regardless of the

leave-one-out validation method used. This overﬁt-

ting on an collected dataset may apply for other re-

search in VAG as well. The acceptance of VAG in

medicine is still far behind the established methods

in this ﬁeld (Shieh et al., 2016). Reasons might be

the lack of resolution, the measurement principle, and

that the sensor attachment on the skin varies by each

individual. Further, the signal transportation is in-

ﬂuenced by various parameters. This indicates that

further research is needed, e.g. optimal sensor place-

ment (Ota et al., 2016), number and type of sensors

(Klemm et al., 2019), weight load during examination

(Andersen et al., 2018) or state of the art feature ex-

traction and classiﬁcation techniques. Even today, a

realistic recognition rate of certain joint diseases like

knee, hip or hand osteoarthritis is not sufﬁcient to ob-

tain a precise result (Altman et al., 1986). Therefore,

we need an enhanced VAG assessment and analysis

system as follows.

Enhancing Vibroarthrography by using Sensor Fusion

131

(a) Scalogram of VAG signal without pathology

(b) Scalogram of VAG signal with pathology

Figure 2: VAG signal comparison.

4 ENHANCED CONCEPT OF THE

VAG SYSTEM

As we worked out in the related work section, the ex-

isting concept of the assessment of a VAG Sensing

System is insufﬁcient. Therefore, we designed a con-

cept that is more advanced, fuses heterogeneous sen-

sors, and compensate the lacks and deﬁciencies of the

existing VAG procedure.

Figure 3: General Architecture of the enhanced VAG sys-

tem.

The major characteristic of the VAG joint analysis is

the trait of dynamic measurement, against still captur-

ing like MRT or X-ray. The VAG measurement can be

performed only in motion. Therefore, we propose and

distinguish between two types of assessment, a fully

controlled movement and the free movement.

The controlled movement is a setting like sitting

in a chair and moving the limbs (Figure 4). The free

movement describes the assessment during activities

like walking, stair climbing, repetitive activities (e.g.

cycling), or other activities of the daily living. The

classiﬁcation of free movement activities and their pa-

rameters is much more difﬁcult that under controlled

conditions. That is the reason we emphasis on the

controlled assessment and will consider the mobile

assessments subsequently in the future. The station-

ary concept (Figure 3) includes the usage of estab-

lished sensors and additional sensors to complete the

setting for a comprehensive assessment. The manda-

tory sensors for an enhanced vibroarthrography are:

• Ultrasound:

Acoustic emission from human knee joints

indicates, that healthy and unhealthy knees can be

successfully distinguished during the sit to stand

movement (Shark et al., 2010).

• Accelerometer:

Assessing the ROM of the joint, detection of

crepitus and other click and crack sounds (Toreyin

et al., 2016).

These sensors will be extended by a net of additional

sensors to improve the signal quality, to reduce the

noise, to enrich the information, and to combine val-

ues with other parameters, e.g. vibration in relation to

joint angle and angular velocity.

4.1 Improvements

The proposed concept includes several improvements

as follows:

• Frequency Band

The movements of the joint causes oscillations

in various frequencies and intensities. Slow os-

cillations can be easily measured with accelera-

tion sensors, higher frequencies are accessible by

microphones. The proposed concept combines

the two sensor types to be able to record the full

range of frequencies. To ensure a reliable over-

lap, we propose to use the acceleration sensor in

full range of sampling rate (e.g. up to 3.2 kHz

by AXDL345) and the microphone (e.g. 20Hz to

44kHz).

SENSORNETS 2020 - 9th International Conference on Sensor Networks

132

• 3D Assessment

Referring to the existing VAG database by Krish-

nan et al., our sensor system consists of multiple

sensors with multiple axes, in comparison to only

a one dimensional acceleration sensor.

Furthermore, we will include the trajectory of the

angle, to recognize a straight movement, torn liga-

ment, or misalignment. Therefore, the integration

of a radar sensor will be an option if the single use

of acceleration sensors is not sufﬁcient.

• Sensor Placement

The coupling of the sensors to the joint is a chal-

lenge. In relation to ultrasonic assessment, a gel

and cuff can be used to enhance the vibration sig-

nal ﬂow. Furthermore, the vibration of a joint are

transmitted also to the bones of the limb. There-

fore, we will examine the signal transmission to

the end of the related bones. We expect a leverage

effect, that the vibrations of the e.g. knee are am-

pliﬁed at the foot angle. Similar to (Klemm et al.,

2019) we may use extensions to enhance the vi-

bration signals.

• Additional Sensors

The concept of an enhances sensor network for

VAG measurements includes additional sensors

to examine supplementary effects, e.g. the mo-

bility and strength of legs with unhealthy and

healthy knees. To distinguish between joint dis-

ease and skeletal muscle disorder, the concept in-

cludes electromyography (EMG) for the measure-

ment of muscle activity (Hollander et al., 2018)

and PPG sensors to determine veins insufﬁciency

by the light reﬂection rheography (LRR).

• Correlation to the Movement Angle

The VAG database by Krishnan et al. contains

only vibration data over time without any correla-

tion of the angle of the joint. Our concept includes

the simultaneously assessment of the joint angle.

• Joint Load

We assume that the load free movement of the

joint generates other vibration than a joint under

stress, as proposed by (Andersen et al., 2018).

Therefore, we designed a load measuring of the

joint to receive a multi-parameter dataset.

• Execution Speed and Repetition

The vibration of the joint differs by the execution

speed of the joint motion (Kernohan et al., 1990).

Therefore, we assess the vibration during certain

motion speed. Currently, we are not aware if vi-

bration vary in repetitions but this is a hypothesis

that has to be examined in future work.

• Machine Learning and Artiﬁcial Intelligence

The core of the VAG Sensing System bases on a

classiﬁer that consists of a neural network. Cur-

rently we propose a convolutional neural network

(CNN) that enables an analysis of complex mo-

tions even during daily activities.

The improvements are combined into a sensor net-

work with a multidimensional signal assessing and

analysis platform.

Figure 4: Heterogeneous sensors as a Sensor Network.

5 DISCUSSION

Signal analysis of mechanical bearings are state of the

art for wear and abrasion estimation. The condition of

human joints are currently only by interest when they

hurt or if a major decline of mobility occurs. Since

now, MRT, CT or ultrasonic examinations are pro-

cedures that are performed whenever a joint diagno-

sis is needed. These methods are expensive and pro-

vide only information recorded in a motionless state.

In contrast, a dynamic assessment is an option to be

used for further diagnostic. The VAG is a dynamic

assessment tool that has the potential to become a

very cheap diagnosis tool because the sensor data as-

sessment is easy to perform, except the data analysis.

Therefore, newest technological concepts in Machine

Learning may be used for a powerful and reliable di-

agnostic. Unfortunately, classiﬁers based on convo-

lutional or recurrent neural networks require a large

dataset that does not exist so far. Existing data bases

do not contain radiation related acceleration data and

mixing all kinds of knee diseases together. For fur-

ther research, we developed the concept for a com-

prehensive data assessment of VAG under the usage

of a sensor network. With this concept, we are able

to establish a comprehensive database that gives the

basis for neural network classiﬁers.

We assume that VAG still has some limitations

but we expect that the advantages outweigh the disad-

vantages. On one side, we are facing technical chal-

lenges like synchronization, easy handling for the pa-

tient with in a setting of self-assessment in the home

Enhancing Vibroarthrography by using Sensor Fusion

133

environment. On the other side, VAG is an indirect

measuring technique. We are not able to determine

the thickness of a cartilage, but we measure the crepi-

tus intensity. Therefore, it will be difﬁcult to achieve

a reliable relation between thickness and sound or de-

gree of disorder or injury. VAG can be used as a gate-

keeper technology and it is convenient to be used for

a long term usage to obtain trends and progress states.

6 CONCLUSION AND FUTURE

WORK

In the paper, we describe a new concept of using the

technology of vibroarthrography (VAG) by using sta-

tionary and mobile assessment of vibrations of human

joints during motion. Hereby, we describe the gen-

eral concept of a stationary assessment system and

outline the improvements. The analysis of the vibra-

tion pattern assessed with the enhanced VAG system

enables a high sophisticate classiﬁcation with neu-

ral nets and the discrimination of healthy or injured

joints. Therefore, we propose to build up a compre-

hensive database, consisting of heterogeneous sensor

data assessed by the enhanced VAG sensor network.

The future work will be the application of the con-

cept and the implementation of a database. Further-

more, we will investigate the relevance of trajectories

of the leg and the interplay of muscle strength, ve-

nous insufﬁciency, and joint disease. VAG did not

found the respected dissemination or usage as a diag-

nosis tool so far, but we assume that the advantage of

a harmless, easy to perform and cheap analysis leads

to its establishment. We propose that not only injured

but also artiﬁcial joints can be analyzed.

ACKNOWLEDGMENTS

This work receives funding from the German Federal

Ministry for Economic Affairs and Energy by ZIM-

16KN04913, related to the project MOREBA.

REFERENCES

Abbott, S. C. and Cole, M. D. (2013). Vibration arthrom-

etry: a critical review. Critical reviews in biomedical

engineering, 41(3):223–242.

Altman, R., Asch, E., Bloch, D., Bole, G., Borenstein, D.,

Brandt, K., Christy, W., Cooke, T., Greenwald, R.,

Hochberg, M., et al. (1986). Development of crite-

ria for the classiﬁcation and reporting of osteoarthri-

tis: classiﬁcation of osteoarthritis of the knee. Arthri-

tis & Rheumatism: Ofﬁcial Journal of the American

College of Rheumatology, 29(8):1039–1049.

Andersen, R. E., Arendt-Nielsen, L., and Madeleine, P.

(2018). Knee joint vibroarthrography of asymp-

tomatic subjects during loaded ﬂexion-extension

movements. Medical & biological engineering &

computing, 56(12):2301–2312.

Athavale, Y. and Krishnan, S. (2019). A telehealth system

framework for assessing knee-joint conditions using

vibroarthrographic signals. Biomedical Signal Pro-

cessing and Control.

Chu, M., Gradisar, I., and Mostardi, R. (1978). A nonin-

vasive electroacoustical evaluation technique of carti-

lage damage in pathological knee joint. Electronics

Letters, 16(4):437–442.

Chu, M. L., Gradisar, I. A., Railey, M. R., and Bowling,

G. F. (1976). Detection of knee joint diseases using

acoustical pattern recognition technique. Journal of

Biomechanics, 9(3):111–114.

Frank, C. B., Rangayyan, R. M., and Bell, G. D. (1990).

Analysis of knee joint sound signals for non-invasive

diagnosis of cartilage pathology. IEEE Engineering in

Medicine and Biology Magazine, 9(1):65–68.

Hollander, D. B., Yoshida, S., Tiwari, U., Saladino, A.,

Nguyen, M., Boudreaux, B., and Hadley, B. (2018).

Dynamic analysis of vibration, muscle ﬁring, and

force as a novel model for non-invasive assessment

of joint disruption in the knee: A multiple case report.

The Open Neuroimaging Journal, 12(1).

Hueter, C. (1883). Grundriss der chirurgie. FCW Vogel.

Kernohan, W. G., Beverland, D. E., McCoy, G. F., Hamil-

ton, A., Watson, P., and Mollan, R. (1990). Vibration

arthrometry. a preview. Acta orthopaedica Scandinav-

ica, 61(1):70–79.

Kernohan, W. G., Beverland, D. E., McCoy, G. F., Shaw,

S. N., Wallace, R. G., McCullagh, G. C., and Mollan,

R. A. (1986). The diagnostic potential of vibration

arthrography. Clinical orthopaedics and related re-

search, (210):106–112.

Kim, K. S., Seo, J. H., Kang, J. U., and Song, C. G.

(2009). An enhanced algorithm for knee joint sound

classiﬁcation using feature extraction based on time-

frequency analysis. Computer methods and programs

in biomedicine, 94(2):198–206.

Klemm, L., S

uhn, T., Spiller, M., Illanes, A., Boese, A.,

and Friebe, M. (2019). Improved acquisition of vi-

broarthrographic signals of the knee joint.

McCauley, T. R., Kier, R., Lynch, K. J., and Jokl, P. (1992).

Chondromalacia patellae: diagnosis with mr imaging.

AJR. American journal of roentgenology, 158(1):101–

105.

McCoy, G. F., McCrea, J. D., Beverland, D. E., Kernohan,

W. G., and Mollan, R. (1987). Vibration arthrography

as a diagnostic aid in diseases of the knee. a prelim-

inary report. The Journal of bone and joint surgery.

British volume, 69(2):288–293.

Msayib, Y., Gaydecki, P., Callaghan, M., Dale, N., and Is-

mail, S. (2017). An intelligent remote monitoring sys-

tem for total knee arthroplasty patients. Journal of

medical systems, 41(6):90.

Murphy, L., Cisternas, M., Pasta, D., Helmick, C., and

Yelin, E. (2017). Medical expenditures and earnings

SENSORNETS 2020 - 9th International Conference on Sensor Networks

134

losses among us adults with arthritis in 2013. Arthritis

Care Res.

Nalband, S., Prince, A., and Agrawal, A. (2017). Entropy-

based feature extraction and classiﬁcation of vi-

broarthographic signal using complete ensemble em-

pirical mode decomposition with adaptive noise. IET

Science, Measurement & Technology, 12(3):350–359.

Nalband, S., Sreekrishna, R., and Prince, A. A. (2016).

Analysis of knee joint vibration signals using en-

semble empirical mode decomposition. Procedia

Computer Science, 89:820 – 827. Twelfth Inter-

national Conference on Communication Networks,

ICCN 2016, August 19– 21, 2016, Bangalore, India

Twelfth International Conference on Data Mining and

Warehousing, ICDMW 2016, August 19-21, 2016,

Bangalore, India Twelfth International Conference on

Image and Signal Processing, ICISP 2016, August 19-

21, 2016, Bangalore, India.

Orhan, S., Akt

urk, N., and Celik, V. (2006). Vibration mon-

itoring for defect diagnosis of rolling element bearings

as a predictive maintenance tool: Comprehensive case

studies. Ndt & E International, 39(4):293–298.

Ota, S., Ando, A., Tozawa, Y., Nakamura, T., Okamoto, S.,

Sakai, T., and Hase, K. (2016). Preliminary study of

optimal measurement location on vibroarthrography

for classiﬁcation of patients with knee osteoarthritis.

Journal of Physical Therapy Science, 28:2904–2908.

Peat, G., Thomas, E., Duncan, R., Wood, L., Hay, E., and

Croft, P. (2006). Clinical classiﬁcation criteria for

knee osteoarthritis: performance in the general pop-

ulation and primary care. Annals of the rheumatic dis-

eases, 65(10):1363–1367.

Pihlajam

aki, H. K., Kuikka, P.-I., Lepp

anen, V.-V., Kiuru,

M. J., and Mattila, V. M. (2010). Reliability of clinical

ﬁndings and magnetic resonance imaging for the di-

agnosis of chondromalacia patellae. JBJS, 92(4):927–

934.

Prior, J., Mascaro, B., Shark, L.-K., Stockdale, J., Selfe, J.,

Bury, R., Cole, P., and Goodacre, J. A. (2010). Anal-

ysis of high frequency acoustic emission signals as a

new approach for assessing knee osteoarthritis. An-

nals of the rheumatic diseases, 69(5):929–930.

Rangayyan, R. M., Krishnan, S., Bell, G. D., Frank, C. B.,

and Ladly, K. O. (1997). Parametric representation

and screening of knee joint vibroarthrographic sig-

nals. IEEE Transactions on Biomedical Engineering,

44(11):1068–1074.

Rangayyan, R. M., Oloumi, F., Wu, Y., and Cai, S. (2013).

Fractal analysis of knee-joint vibroarthrographic sig-

nals via power spectral analysis. Biomedical Signal

Processing and Control, 8(1):23–29.

Shark, L.-K., Chen, H., and Goodacre, J. (2010). Discover-

ing differences in acoustic emission between healthy

and osteoarthritic knees using a four-phase model of

sit-stand-sit movements. The open medical informat-

ics journal, 4:116–125.

Shieh, C.-S., Tseng, C.-D., Chang, L.-Y., Lin, W.-C., Wu,

L.-F., Wang, H.-Y., Chao, P.-J., Chiu, C.-L., and Lee,

T.-F. (2016). Synthesis of vibroarthrographic signals

in knee osteoarthritis diagnosis training. BMC re-

search notes, 9(1):352.

Tai, S. M., Munir, S., Walter, W. L., Pearce, S. J., Wal-

ter, W. K., and Zicat, B. A. (2015). Squeaking

in large diameter ceramic-on-ceramic bearings in to-

tal hip arthroplasty. The Journal of arthroplasty,

30(2):282–285.

Tandon, N. and Choudhury, A. (1999). A review of vibra-

tion and acoustic measurement methods for the detec-

tion of defects in rolling element bearings. Tribology

international, 32(8):469–480.

Teague, C. N., Hersek, S., Toreyin, H., Millard-Stafford,

M. L., Jones, M. L., Kogler, G. F., Sawka, M. N., and

Inan, O. T. (2016). Novel methods for sensing acous-

tical emissions from the knee for wearable joint health

assessment. IEEE transactions on bio-medical engi-

neering, 63(8):1581–1590.

Toreyin, H., Jeong, H. K., Hersek, S., Teague, C. N., and

Inan, O. T. (2016). Quantifying the consistency of

wearable knee acoustical emission measurements dur-

ing complex motions. IEEE journal of biomedical and

health informatics, 20(5):1265–1272.

Wu, Y. and Krishnan, S. (2011). Combining least-squares

support vector machines for classiﬁcation of biomed-

ical signals: a case study with knee-joint vibroarthro-

graphic signals. Journal of Experimental & Theoreti-

cal Artiﬁcial Intelligence, 23(1):63–77.

Enhancing Vibroarthrography by using Sensor Fusion

135