Evaluation of a Motion Simulation Platform for Vestibulo ...742143/FULLTEXT01.pdf · 4.1 Sensors, Signal Processing and Analysis Software . . . . . . .15 ... research takes place

i

Evaluation of a Motion Simulation Platform for

Vestibulo-Ocular Research

SIMON LUNDBERG

Master of Science Thesis in Medical Engineering Stockholm 2014

II

This master thesis project was performed in collaboration with the Department of Clinical Neurosciences, Karolinska Institutet

Supervisor at Karolinska Institutet: Assoc. Prof. Tony Pansell

Evaluation of a Motion Simulation Platform for Vestibulo-Ocular Research

Utvärdering av en rörelsesimuleringsplatform

för vestibulo-okulär forskning

S I M O N L U N D B E R G

Master of Science Thesis in Medical Engineering Advanced level (second cycle), 30 credits

Supervisor at KTH: Chiara Giordano Examiner: Svein Kleiven

School of Technology and Health TRITA-STH. EX 2014:78

Royal Institute of Technology KTH STH

SE-141 86 Flemingsberg, Sweden http://www.kth.se/sth

III

IV

Abstract

The vestibuloocular reflex can be manually elicited by tilting or rotatingthe head. Manual techniques serve their purpose well and is the goldenstandard in the clinical work, but they lack control of velocity and movementpattern. However, motion simulation platforms enable automatic control ofboth velocity and movement pattern.

One motion simulation platform, named BIRGIT, has been built at theDepartment of Clinical Neurosciences at Karolinska Institutet but has notyet been in service and require a performance evaluation.

The objectives with this thesis is to evaluate the accuracy and precisionof BIRGIT and evaluate how bodyweight and movement direction impactthe performance. The thesis also evaluate whether it is possible to stabilizethe head of the patient during the acceleration phase.

Repeated measurements of acceleration with different loads, desired ac-celeration, direction and motion type (rotational and translational), are per-formed. Dummies are used to simulate bodyweight in the performance studyand real persons are used in the head stabilization study. Analysis of variance(ANOVA) is the main statistical tool.

The results suggest that the platform does not perform equally at dif-ferent load or directions and that there is a bias between desired and trueacceleration. The main problems are an inclination of the rails, upon whichthe chair is mounted, that causes differences between directions and an un-desirable performance characteristic for rotational motions.

The stabilization study suggest that the head can be stabilized.

Keywords: Vestibuloocular reflex, motion simulation platform, head sta-bilization, BIRGIT

V

VI

Sammanfattning

Den vestibulo-okulara reflexen kan framkallas genom att manuellt rotera ellerrycka pa huvudet. Att manuellt framkalla reflexen fungerar i de flesta sam-manhang val och ar standard i manga undersokningar. Dock gar det inte atttill fullo kontrollera vare sig hastighet eller rorelsebana perfekt. Emellertidgar detta att kontrollera genom att anvanda sig av en sa kallad rorelsesimu-leringsplatform.

En sadan platform, dopt till BIRGIT, har byggts vid Institutionen forkliniska neurovetenskaper vid Karolinska Institutet. Denna har ej borjatanvandas annu da dess prestanda forst behover utvarderas.

Syftet med denna uppsats ar att utvardera precision och noggrannhethos BIRGIT. Dessutom, att utvardera hur kroppsvikt och rorelseriktninginverkar pa prestandan. I arbetet ingar aven att testa om det ar mojligt attstabilisera huvudet under accelerationsfasen.

Repeterade matningar av sann acceleration med olika last, riktning, borac-celeration och rorelsetyp (rotation eller sidledes) genomfordes. Testdockoranvandes for att simulera lasten i prestandatesterna och riktiga testpersoneranvandes i huvudstabiliseringsdelen. Variansanalys (ANOVA) var det hu-vudsakliga statistiska verktyget.

Resultatet antyder att last och rorelseriktning inverkar pa acceleratio-nen och prestandan och att den sanna accelerationen alltigenom ar lagre anboraccelerationen. Det finns tva storre problem hos plattformen, det forstaar att upphangningen till stolen lutar och detta resulterar i en skillnad mellanriktningarna (det gar fortare nedfors). Det andra problemet ar ett cyklisktuppforande nar flera rotationsrorelser skall folja pa varandra.

Stabiliseringsstudien visade att det gar att stabilisera huvudet.

Nyckelord: Vestibulo-okulara reflexen, rorelsesimuleringsplatform, hu-vudstabilisering, BIRGIT

VII

VIII

Contents

List of Figures XI

List of Tables XIII

List of Abbreviations XV

1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Demarcations . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Theoretical Background 32.1 The Vestibular System . . . . . . . . . . . . . . . . . . . . . . 3

2.1.1 Balance and the Vestibular System . . . . . . . . . . . 32.1.2 The Vestibuloocular Reflex . . . . . . . . . . . . . . . . 5

2.2 Assessments of Eye Movement . . . . . . . . . . . . . . . . . . 52.3 Elicitation of the Vestibuloocular Reflex . . . . . . . . . . . . 7

2.3.1 Manual Elicitation . . . . . . . . . . . . . . . . . . . . 72.3.2 Thermal Elicitation . . . . . . . . . . . . . . . . . . . . 72.3.3 Vestibular Autorotation Test . . . . . . . . . . . . . . . 82.3.4 Motion Simulation Platforms . . . . . . . . . . . . . . 8

3 BIRGIT 113.1 Mechanical Parts and the Motors . . . . . . . . . . . . . . . . 113.2 Drive Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.3 Software and Controlling . . . . . . . . . . . . . . . . . . . . . 123.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4 Method 154.1 Sensors, Signal Processing and Analysis Software . . . . . . . 154.2 Study I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.2.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . 16

IX

4.2.2 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 174.3 Study II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4.3.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . 184.3.2 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 18

4.4 Study III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184.4.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . 194.4.2 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 19

5 Results 235.1 Study I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

5.1.1 Translational Motion ANOVA . . . . . . . . . . . . . . 235.1.2 Rotational Motion ANOVA . . . . . . . . . . . . . . . 31

5.2 Study II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335.3 Study III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

5.3.1 Acceleration of the Head . . . . . . . . . . . . . . . . . 38

6 Discussion 396.1 General Discussion . . . . . . . . . . . . . . . . . . . . . . . . 39

6.1.1 Study I . . . . . . . . . . . . . . . . . . . . . . . . . . 396.1.2 Study II . . . . . . . . . . . . . . . . . . . . . . . . . . 406.1.3 Study III . . . . . . . . . . . . . . . . . . . . . . . . . 41

6.2 Level of Significance . . . . . . . . . . . . . . . . . . . . . . . 426.3 Limitations and Suggestions . . . . . . . . . . . . . . . . . . . 42

7 Conclusions 45

8 Acknowledgements 47

Bibliography 49

A The Dummies 53

B Post hoc tests, study I translational 55

C Post hoc tests, study I rotational 59

D Post hoc tests, study II 63

E Post hoc tests, study III 65

X

List of Figures

2.1 Details over the vestibular system . . . . . . . . . . . . . . . . 42.2 The vestibuloocular reflex . . . . . . . . . . . . . . . . . . . . 62.3 Barany chair . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1 Overview of BIRGIT . . . . . . . . . . . . . . . . . . . . . . . 12

4.1 Degrees of freedom of the sensors . . . . . . . . . . . . . . . . 154.2 Stabilizing device . . . . . . . . . . . . . . . . . . . . . . . . . 204.3 Measured properties in study III . . . . . . . . . . . . . . . . . 21

5.1 Left translational measurements . . . . . . . . . . . . . . . . . 245.2 Right translational measurements . . . . . . . . . . . . . . . . 245.3 Clockwise rotational measurements . . . . . . . . . . . . . . . 255.4 Counter-clockwise rotational measurements . . . . . . . . . . . 255.5 Translational ANOVA acceleration . . . . . . . . . . . . . . . 275.6 Translational ANOVA weights . . . . . . . . . . . . . . . . . . 285.7 Translational ANOVA weights*acceleration . . . . . . . . . . . 295.8 Translational ANOVA direction . . . . . . . . . . . . . . . . . 295.9 Translational ANOVA direction*acceleration . . . . . . . . . . 305.10 Translational ANOVA weight*direction . . . . . . . . . . . . . 305.11 Rotational ANOVA acceleration . . . . . . . . . . . . . . . . . 325.12 Rotational ANOVA weights . . . . . . . . . . . . . . . . . . . 325.13 Rotational ANOVA weights*acceleration. . . . . . . . . . . . . 335.14 Rotational ANOVA direction. . . . . . . . . . . . . . . . . . . 345.15 Rotational ANOVA direction*acceleration. . . . . . . . . . . . 345.16 Rotational ANOVA weight*direction. . . . . . . . . . . . . . . 355.17 Within set ANOVA . . . . . . . . . . . . . . . . . . . . . . . . 355.18 Between set ANOVA . . . . . . . . . . . . . . . . . . . . . . . 375.19 Effect of stabilizing device . . . . . . . . . . . . . . . . . . . . 37

A.1 Weight distribution of dummies . . . . . . . . . . . . . . . . . 54

XI

XII

List of Tables

5.1 Mean value and variance . . . . . . . . . . . . . . . . . . . . . 265.2 ANOVA, translational motions . . . . . . . . . . . . . . . . . . 275.3 ANOVA, rotational motions . . . . . . . . . . . . . . . . . . . 315.4 ANOVA, within set . . . . . . . . . . . . . . . . . . . . . . . . 365.5 ANOVA, between sets . . . . . . . . . . . . . . . . . . . . . . 365.6 ANOVA, stabilization of head . . . . . . . . . . . . . . . . . . 365.7 Head acceleration properties . . . . . . . . . . . . . . . . . . . 38

B.1 Translational, post-hoc test of accelerations . . . . . . . . . . 55B.2 Translational, post-hoc test of weights . . . . . . . . . . . . . 55B.3 Translational, post-hoc test of weight*acceleration . . . . . . . 56B.4 Translational, post-hoc test of direction*acceleration . . . . . 57B.5 Translational, post-hoc test of weight*direction . . . . . . . . 58

C.1 Rotational, post-hoc test of accelerations . . . . . . . . . . . . 59C.2 Rotational, post-hoc test of weights . . . . . . . . . . . . . . . 60C.3 Rotational, post-hoc test of weight*acceleration . . . . . . . . 61C.4 Rotational, post-hoc test of direction*acceleration . . . . . . . 62

D.1 Post-hoc test of within set repetition . . . . . . . . . . . . . . 64

E.1 Post-hoc test of subject*stabilization . . . . . . . . . . . . . . 66

XIII

XIV

List of Abbreviations

ANOVA Analysis of variance

CCW Counter-clockwise

CG Center of gravity

CSN Central nervous system

CW Clockwise

DF Degrees of freedom

HIT Head impulse test

HSD Honest significant difference

MS Mean of squares

MSP Motion simulation platform

OCR Ocular counter-rolling

SNR Signal to noise ratio

SS Sum of squares

VAT Vestibular autorotation test

VO Vestibular organ

VOR Vestibuloocular reflex

XV

XVI

1

Introduction

This chapter details the aims of this thesis. It starts with a brief introductionto the balance and vestibular organ. Thereafter the laboratory and its motionsimulation platform are presented, followed by what questions the thesis aimsto answer.

1.1 Background

We humans possess a complex and delicate sensory system to maintain bal-ance. It can roughly be separated into three parts: the proprioceptive, thevisual and the vestibular system. The vestibular organ (VO) is situated inthe inner ear and senses head accelerations. It consist of the utricle and sac-cule which senses translational acceleration and the three semicircular canalswhich can sense rotational acceleration. This detection of movement aids uswhen we walk, run, stand still and sit down.[1]

The VO does not solely help us to keep balance but it also influencesvision. To see clear we must keep the image, falling on the retina, still. If thehead is moved in any direction the picture would move on the retina, this iswhere the VOR sets in. The vesitbuloocular reflex (VOR) is responsible tofixate on the object of interest and hence keeping the picture on the retinastill. When head movement is sensed by the VO, vestibular signals is sentto the eyes muscles to move the eye in opposite direction to compensate thehead movement. The reflex is active at all time and compensate for horizon-tal, vertical and torsional (rotation around visual axis) head movement andcombinations of these.[2]

At the Department for Clinical Neuroscience at Karolinska Institutet,research takes place at the Bernadotte laboratories for pediatric ophthal-mological research. To accurately evaluate the VOR, a motion simulation

1

platform (MSP) has been constructed. The platform, named BIRGIT, cansimulate translational and rotational motions. For the time being, VOR havebeen manually provoked by tilting and rotating the head. The drawback ofmanual methods is the difficulty of repeating the same movement pattern,and therefore the same head movement, twice. However the platform aimsto automatically control head movements with high repeatability.

1.2 Objectives

The MSP has not been in service yet; in order to be, it has to be evaluatedand tested. The evaluation includes repeated measurement of accelerationi.e. precision and accuracy, how bodyweight impacts the performance and ifthe chair perform the same in both directions.

When a subject is exposed to acceleration the head will, due to its inertia,lag behind the rest of the body (where the force is acting). The head has tobe stabilized to follow the chair movement as precise as possible. The secondpart of the thesis evaluate if this is possible.

Two hypotheses are tested in the evaluations:

H01 The patients bodyweight does not impact the MSPs behavior

H02 The head can not be mechanically stabilized

1.3 Demarcations

The thesis includes a chapter covering the physiological properties of theVOR and an introduction of examination methods of this ocular motor reflex.The method and the result are described for each sub-study. In the end thereis a discussion based on the results together with conclusions.

The goal is not to improve BIRGIT or fix artifacts that are found, but togive a foundation for future usage and improvements of the platform. Sincethe platform is unique it is difficult to compare it to other MSPs with perhapsdifferent intended use or other motion pattern.

2

2

Theoretical Background

This chapter details background information to the areas of vestibular re-search covered by this thesis. It starts by describing important physiologyfeatures followed by examination methods and ends with a description ofmotion simulation platforms.

2.1 The Vestibular System

This section gives an introduction to the anatomy and physiology of the bal-ance organ in the inner ear. It is followed by an explanation of the vestibu-loocular reflex and how humans are able to keep focus while moving the head.At last there is a small part about vestibular disorders.

2.1.1 Balance and the Vestibular System

The brain has three main sources of information that contributes to balance:visual, proprioceptive and vestibular input. Normally, information from theseare non-conflicting and helps us to keep balance in every day situations. If theinformation is conflicting, the vestibular input tends to have highest priority.because it relies on gravity which direction is constant and thus can be seenas an unchanged reference. This in contrast to the proprioceptive systemwhich is based on for example the relative position of the different body partand vision, which lack an objective reference point.[3]

The vestibular organ (VO), seen in Figure 2.1a, is located in the labyrinthof the inner ear next to the cochlea. The VO can be described like a liquidfilled system of tubes and compartments which comprises a total of five sen-sory sites divided into two major parts; the otholitic organs and the semi-

3

(a) The inner ear with the VOto the left

(b) The ampulla with arrows in-dicating direction of motion

Figure 2.1: Anatomical details.Source: http://what-when-how.com/neuroscience/auditory-and-

vestibular-systems-sensory-system-part-2/

circular canals. All sensory sites are sensitive to specific motion of the headand the otholites care for linear head motions and the canals for rotationalhead motions.[3]

The Otholitic Structure

Two of the VOs five sensory sites are in the otholitic structure. The first isthe utricle which senses accelerations in the horizontal plane. Above a layerof sensory cells, it lies a gelatinous membrane covered with calcium carbonatecrystals called otoconia. They have a relatively high density compared to thegelatinous membrane and the relatively heavy otoconias is only connectedto the stiff surrounding via his membrane. When the head is acceleratingthe otoconias will lag behind the surrounding structure and this displacementwill be sensed by the sensory cells and interpreted as a horizontal accelerationof the head. The other sensory site in the otholitic structure is the sacculewhich, roughly, can be seen as a vertical orientated utricle and consequentlysenses vertical motions.[3]

The Semicircular Canals

The three remaining sensory sites, of the VO, are in the semicircular canals.These liquid filled canals are almost perpendicular to each other and sensesrotational movement in their respective direction. Every canal has a smallwidening called the ampulla where the sensory cells are located. They aresituated in a structure called cristae ampullaris upon which a gelatinous sailcover a cross-section of the ampulla. When the head is rotated the liquid in

4

the canal will lag behind the surrounding structure and deform the sail.This deformation is interpreted as a rotation of the head. See Figure 2.1bfor major details on the ampulla.[3]

2.1.2 The Vestibuloocular Reflex

To see clear, the eye must be able to keep an image still on the retina. Iffocusing on an object we must be able to compensate for both large movementand small disturbances for example the head movement when walking. Thereis one major mechanism that aids us to stabilize the picture on the retina,the vestibuloocular reflex.[3]

If we focus on an object and then turn the head, the picture of the objectwill rapidly come out of focus of the eye. The motion of the head is sensedby the VO and a signal is sent to the extraocular muscles to react and movethe eyes in opposite direction so that the image is stabilized during the headmovement. Figure 2.2 illustrates the reflex. Halmagyi et al. [4] measured thedelay of the reflex and found it to be around 20 ms and sometimes as low as10 ms; the reflex has one of the shortest latencies in the human body.

The VOR aids to counteract head movement and stabilize the vision notonly vertically and horizontally. If the head is tilted towards the shoulderthe eyes will rotate around their visual axis in counter-direction; a phenom-ena called ocular torsion, and the reflex is denoted ocular counter-rolling(OCR).[5] VOR refers to both VOR and OCR in this thesis. To actuate thesteering signals from the VO, there are extra-ocular muscles serving for themotion of the eyes. A fast simultaneous movement of both eyes - in the samedirection - is called a saccade. [3, 6]

2.2 Assessments of Eye Movement

Since the VO drives the VOR-response it is possible to identify diseasesand malfunctions, such as vertigo, by examining the VOR.[3, 6] To do sothe VOR must be stimulated and the eye movement measured. The easiestway to assess the movement is through direct gross visual observations; amethod that requires the examiner to be proficient enough. There are other,quantitative, methods to assess the movement and two of them representingdifferent paradigms in how to assess, are described here.[7]

5

Figure 2.2: The vestibu-loocular reflex. The rings atthe bottom is the semicircularcanals. Clockwise rotationof the head causes the liquidinside the canal to relativelymove counter clockwise. Therotation is detected and sig-nals are sent to the extraocularmuscles to react and movethe eyes counter clockwisein order to keep the imageon the retina still. Source:http://en.wikipedia.org/wiki/File:ThreeNeuronArc.png

Scleral Search Coil

Scleral search coil represent mechanical on-eye assessment technique.

A fine coil is embedded in to a contact lens and placed on the eye of thepatient. The patient is seated inside a frame surrounded by electromagneticcoils. The external coils alternate their activity and the continuously chang-ing magnetic field will induce a signal in the lens. The alternating schemeis known so it is possible to translate the signal from the lens in to position;it is possible to detect horizontal, vertical and torsional movement with thistechnique.[2, 6]

Video Recording

Video recording represent eye tracking by using light.

Video recording of eye movement enable offline examination. High defini-tion cameras and modern image processing software enable to detect move-ment which is either too small or too rapid for the naked eye to detect.[6]

Visible or infrared light can be used as light source. Infrared light sourceis used in video-oculography where a helmet with cameras and light sourcesare mounted on the patients head. Infrared light enables recording in a darkroom or when lightning is a variable in the measurements e.g. eye trackingin different lightnings.[2, 6]

6

2.3 Elicitation of the Vestibuloocular Reflex

This section introduces techniques to stimulate the VO and trigger the vestibu-loocular reflex. It starts with simple manual techniques and proceeds bydescribing mechanical passive and active head motions. In the end motionsimulation platforms and their advantages are described. The sections aimsto explain and give a background to why such platforms are used.

2.3.1 Manual Elicitation

Halmagyi and Curthoys [8] have introduced the head impulse test (HIT) toassess the function of the VO. The examiner quickly moves the patients headto one side as the patient focus on the examiners eye. This head movement isdenoted as passive and the examiner visually observe if the patients eye losesfixation. A healthy patient should be able to maintain gaze on the exam-iner without any compensatory fixating saccades. The test was introducedto detect unilateral loss of semicircular canal function since compensatorysaccades only occur when the head is moved towards the lesioned ear. HIThas been shown to be very useful especially in the emergency unit where fastexamination has to be made, for example to distinguish between vestibularneuritis and vestibular infarction.[9, 10]

One problem with manual elicitation is the difficulties to create puremotions. Sometimes it is desirable to evaluate a single semicircular canalwhich can be done if only one sensory site is activated i.e. the movement mustbe without distortion. Ramat et al. [11] introduced the head sled to minimizerotational and off-axis components of the movement when performing lateralstimuli. The head is fastened in a frame which roll upon two rails.

Today’s methods of manually elicitation of the VOR serve their purposewell. The methods are mostly qualitatively but give useful information tothe examiner and are used for both screening and diagnostic purposes. Mostmethods can be performed without any advanced aid or device and the num-bers of maneuver designed for specific examination is a helpful toolbox atmedical units.[6]

2.3.2 Thermal Elicitation

There are a few ways to test the connection between the VO and the eyes andhence the VOR, without inducing head movement. One example is calorictesting where either warm or cold water is irrigated on one side to change thetemperature in the inner ear. Temperature changes induces nystagmus: theeye beats towards a hot ear and away from a cold ear at a healthy patient.[12]

7

2.3.3 Vestibular Autorotation Test

When performing vestibular autorotation test (VAT), electrodes are attachedaround the eyes. What distinguishes this method from the previous one isan active eliciting of VOR induced through head movement. The subjectis asked to shake the head vertically (as in yes) or horizontally (as in no)at the same pace as an electrical metronome dictates; the metronome con-tinually increases the pace. Accelerometers and velocity sensor are attachedon the head to keep track of the head movement and the eye movement isrecorded.[3]

The method can be seen as a controlled active VOR elicitation methodbut it still lacks the pure movement seen when using the sled seen in subsec-tion 2.3.1. It has also been shown that the test-retest repeatability can bepoor [13] and some patient have difficulties to perform the test so the resultshould be interpreted with some carefulness. Another problem is vestibularadaption, i.e. the patient learns how to compensate and the validity of thetest hence decrease.[14, 15]

2.3.4 Motion Simulation Platforms

Motion simulation platforms (MSP) aim to generate a very precise movementwith high accuracy and repeatability. Most MSPs can only perform oneor a few types of movement e.g. rotate around the vertical axis or movelinearly forward and backward but platforms can have up to six degrees offreedom (DF). The platform often have a chair where the patient is seatedand fastened, the movement is passive and the whole body is moved.[3]

One type of MSP is the rotary chairs, sometimes denoted vestibular chair.The first such chair is the Barany chair [16], seen in Figure 2.3, which is a1DF (rotation around vertical axis) platform. The chair is an advanced swivelchair with an arch around it for safety and for the examiner to manuallycontrol speed and acceleration.

However, velocity, acceleration or position is not automatically controlled.This could be achieved via computer controlled and motorized systems. Ifthe chair is working autonomous it is for example possible to enclose it ina dark room or compartment and use the advantages given by e.g. infraredvideo recording (see section 2.2). The MSP also enable to re-test the verysame stimulation at re-visits.

Studies have been done to evaluate whether it is better to use a rotarychair than manual passive and active techniques. Maes et al. [17] have testedthe test-retest reliability for rotational chairs and concluded it to be a highlyreliable technique. Goebel et al. [18] compare rotary chairs with both pas-

8

Figure 2.3: A Barany chair during an examination.Source: U.S. Air Force, picture is in the public domain

sive and active manual whole body rotation and head on body rotationaltechniques. They found that manual whole body rotation techniques wascomparable with rotary chairs in the range 0.025-1 Hz and that passive andactive head on body techniques showed identical result in the same range.[3]

Another simple motion simulator is a sled mounted on rails. It sharesmost characteristics with the Barany chair except that it moves linear insteadof rotating. One problem that occurs in the sled, as well as in other MSP,is the risk that the head might not perfectly follow the bodys movement.There are some solutions to this problem e.g. support cushions which is whatNational Space Biomedical Research Institute [19] use in their sled. Exampleof such sled is the one by Ramat et al. [11] mentioned in subsection 2.3.1.

There are also more advanced MSP; Dits et al. [20] have built a 6DF MSP.The platform consist of a chair that can move translational and rotationalin all directions. The chair can create translational stimulation pulses up to100 deg/s2. To stabilize the head they use support cushions and a dental bitboard to minimize unintended head movement.

9

10

3

BIRGIT

This chapter gives basic information of the motion simulation platform (MSP)BIRGIT. Its mechanical components are presented together with controllingof motion patterns and their limitations.

The motion simulation platform BIRGIT creates a controlled environ-ment in which specific motion patterns can be automatically reproduced;motion simulation thus represent the capability of reproducing desired mo-tions in a controlled manner.

3.1 Mechanical Parts and the Motors

The MSP BIRGIT (Figure 3.1) consists of a chair mounted on two horizontalrails. The upper and lower link point of the chair and the rails are bothattached at the chairs midline and moved individually by one three-phaseelectrical motor each. The lower link point is not fixated at the chair but canslide up and down along the midline. This mobility of the link point enablesthe chair to not only slide sideways but also to rotate. Maximum distanceof motion is 1.44 m along the rails. On the rails there are emergency breaksto turn off the engines when the chair reaches the end point of the movabledistance. In addition there are also homing sensors used to calibrate thechair in to vertical position before each run.[21]

There is a four-point seat belt mounted on the chair to make sure thesubject keep seated in the chair.

By using a laser spirit level it was discovered that both rails have a smallinclination by 0.5 deg. The lower end is to the left as seen by a person seatedin the chair. This inclination was not known before this thesis work.

11

Figure 3.1: Overview of BIRGIT

3.2 Drive Files

The desired motion pattern is specified in a drive file by a numbers of param-eters. Every unique movement has to be specified and the next movementstarts where the last one ended. Before the first motion is initiated, the chairneeds to be calibrated along the vertical position. The parameters to spec-ify are: type of movement, i.e. rotational or translational; the amplitude ofeither velocity or acceleration; end position; the direction, left/clockwise orright/counter-clockwise; any time delay before next movement is actuated;and last if it is a rotational movement, the center of rotation has to be spec-ified. For example, if multiple rotational movement is desired it is necessaryto specify the rotation but also the movement back to vertical position.

When the drive file is loaded in to the program it is necessary to chose ifthe movement shall be optimized for acceleration or amplitude. If accelera-tion is chosen the amplitude will be arbitrary.

3.3 Software and Controlling

The desired movement sequence is sent to the electronic control unit that isthe location of the positioning controller. The signal from the controller is

12

sent forward to the driver which regulates the electrical signal to the motors.The rotation on the axis, inside the motors, is sent as feedback to the con-troller. In this way the controller always should be able to know the positionof the chair. The drive files and control files are written in the language Mint(Motion Intelligence).[22]

3.4 Limitations

There are some safety limitations in the software. First the program checksfor contradictions in the drive file, e.g. it is not possible to move the samelink point both right and left at the same time. If no such errors are found itchecks for positioning errors that is whether any of the link point is calculatedto cross the end points.

This safety control was shown to cause problems for the design of drivefiles with multiple repetitions. When the movement is optimized for acceler-ation, the retardation starts after reaching the desired position. Supposedly,the system loose track of the chair position. Even if manually calibrating thechair back to vertical position and one by one adding a repetition the systemcomplains that the chair is calculated to be out of range. No work-aroundsolution is found to this limitation.

13

14

4

Method

This chapter details methods and devices used in the experiments. The firstsection describes common devices and software used in all studies. Thisis followed by description of methods and statistical tools for every studyindividually. The explanation of the statistics is more thorough in study Iand serves as basis for the other studies.

4.1 Sensors, Signal Processing and Analysis

Software

The sensors used in all the studies come from Chronos Inertial Tracking Sys-tem (C-ITS), Chronos Vision GmbH, and consist of two triaxial sensors; onefor linear acceleration and one for angular rate. Figure 4.1 shows the sixdegree of freedom that the sensor system detects. If nothing else is stated,accelerations and velocities are the resulting vectors (Euclidean). This mini-mize measurement distortion caused by sensors that are not perfectly alignedwith the direction of movement.

A discrete time moving average filter is used with a time window of 50

X

Y

Z

Figure 4.1: Degrees of freedom of the sensors

15

samples. The filter is a Savitzky-Golay filter which Guinon et al. [23] showedis better than a ordinary moving average filter in preserving peak heightand width. It is based on a least squared polynomial fitting instead of localaveraging.[24]

Signal processing is done in OriginPro version 8 from OriginLab R© andstatistical analysis is performed in Statistica 6 from StatSoft R©.

4.2 Study I

In this study the precision and accuracy of BIRGIT is measured. It is followedby an analysis of how direction and load impacts the performance. Below,the set up and statistical methods are described.

4.2.1 Method

The sensors were mounted with clamps on the chair close to the upper linkpoint and the sampling frequency is set to 400 Hz.

The study was divided in to two sub-studies, one for rotational and onefor translational movements. Center of rotation was set to the upper linkpoint in the rotational part.

Three different loads were used in the chair: 0, 40 and 80 kg. Theyare denoted empty, child and adult. Instead of using real test subjects,two dummies were created, see Appendix A for details on dummy design.Drive files were created to run the chair in both directions (left and right fortranslational motions and clockwise (CW) and counter-clockwise (CCW) forrotational motions) and with four accelerations: 50, 100, 150 and 200 cm/s2

or deg/s2.

The number of repetition for every combination of load-direction-accelera-tion was set to 20 after checking convergence of the variance. Due to theconstraints described in section 3.4 it was not possible to design a drive filewith 20 repetition. Instead multiple sets with varying number of repetitionswere run. For all files with translational movement it was possible to createdrive files with 10 repetitions and two such set was run. For a rotationalacceleration of 50 deg/s2, two sets were run with ten repetitions each; at100 and 150 deg/s2 four sets were run with five repetitions each; and at 200deg/s2, five sets were run with four repetitions each.

16

4.2.2 Statistics

In the present study, ANOVA statistical analysis was used. ANOVA is aset of statistical methods that allow to compare and contrast two or moresets of data, comparing the inner variability of these sets with the variabilityin between different sets.[25] The ANOVA in this study had three factors:weight, comprising of the levels empty, child and adult; direction, comprisingthe levels left/CW and right/CCW; and desired acceleration, comprising ofthe levels 50, 100, 150 and 200. The significance level of the ANOVA was setto α = 0.05.

The outcome from the ANOVA includes an F-ratio, which is the variationbetween groups divided by the variation within groups (Equation 4.1). TheF-ratio is compared with a F-critical value derived from an F-distributionand the degrees of freedom in the test. The null-hypothesis can be rejectedif the F-ratio exceeds the critical value.[26]

F =MSbetween

MSwithin

=between-group variability

within-group variability(4.1)

Sphericity refers to the equality of variance of the difference between mea-sured combination of levels and violation of sphericity increase the risk fortype I-error, that is, the test becomes to liberal. The Greenhouse-Geissercorrection can be used to compensate when sphericity is violated by correct-ing the degrees of freedom of the F-distribution in the test.[27] All observa-tions were tested for sphericity with Mauchly’s test and if the assumption ofsphericity was violated, the Greenhouse-Geisser correction was used.

The effect size is a measure of association between the independent anddependent variables.[28] In the present study the effect size is described witheta partial squared, η2partial. This variable is defined as the proportion of thetotal variability attributed to a given factor.

Post hoc tests were performed with Tukey’s honest significance differencetest (Tukey’s HSD-test), also known as Tukey-Kramers’ method. This is amultiple comparison technique to find means that are significantly differentwithin a group. The test is essentially a t-test with the modification that itcompensate for the increased risk of type I-error that repeated t-tests usuallysuffer from.[29]

4.3 Study II

In study I a cyclic behavior was found when rotational drive files were run.The cycles follows the number of repetition in the drive files. This artifactwas further investigated and a study was designed to further evaluate the

17

precision and accuracy. This study aims to investigate the variance withinsets, i.e. between repetitions in one drive file, and between sets, i.e. thedifference when the same drive file was run multiple times.

4.3.1 Method

To investigate whether there are difference within a set the combination ofdirection and acceleration with the smallest variance was picked to representthe observation. A conservative assumption was made such as if significantdifferences between repetitions within this set are found, this should hold forall sets with higher variance. A number of 20 sets were measured with anempty chair, CCW direction and an desired acceleration of 50 deg/s2; thiswas the combination with lowest variance.

Another conservative assumption was made such as if no significant differ-ence between sets are found, this should hold for all sets with lower variance.A number of 20 sets was collected with an empty chair, CW direction and anacceleration of 200 deg/s2; this was the combination with highest variance.

4.3.2 Statistics

Two repeated measures ANOVA was performed with a significance levelof α = 0.05. Observations were tested for sphericity with Mauchly’s test.When the assumption of sphericity was violated this was corrected usingGreenhouse-Geisser correction. To study differences within sets the datawas organized in 10 observations with 20 repetitions each. In the betweenset analysis the data was organized into 20 observations, one for every run.Post hoc testing was performed with Tukeys’ HSD-test and the effect size isdescribed with η2partial and post hoc test performed with Tukeys’ HSD-test.See subsection 4.2.2 for details on statistics.[25, 26, 27, 28, 29]

4.4 Study III

When MSPs are used, the objective is to investigate the stimulus on the VO.However, at high accelerations the neck is incapable to immediately followthe head, due to the head mass (inertia). If the motion of the chair doesnot correspond to the anatomical motion, there is little use of the motionplatform. The stimulation of the VO is indeed not controllable.

Stabilization of the head, or fixation of the head to the chair is needed;this study aims to improve MSP precision and biofidelity of the movement.

18

4.4.1 Method

The more unstable the neck is, the less it will be able to keep up with thechair in the acceleration phase. The rise time, Tr, is the time that it takesto go from 10% to 90% of the final value. If the neck of a subject is totallystiff, the rise time of the chair and the subject’s head will overlap. The moreunstable the head is the longer will the rise time be. Results of Study II areused to set the behavior of the chair and the acceleration features.

The C-ITS sensors were mounted on a Chronos Eye Tracking Devicehelmet (C-ETD) from Chronos Vison GmbH, which is designed to track eyemovement and hence can fit very tightly to the head. Five subjects wereused and they were seated in the chair with the helmet on and asked to focuson a horizontal line five meters ahead. A number of 20 repetition of lefttranslational movement with an acceleration of 150 cm/s2 was run withoutany stabilizing device.

A head side support from Anatomic SITT (article 77145) was mountedto the frame of the chair (Figure 4.2) and adjusted to fit tight to the cheekof the subject. Another 20 repetitions were run with the stabilizing device.

All five subjects were men with no known (diagnosed) instability in theirneck. Their average weight was 81,6 kg with a standard variation of ±6 kg.They participated voluntarily and were informed about the risks and aboutthe study.

4.4.2 Statistics

The primary variable was the rise time of acceleration, which is proportionalto the stability of the head. Two secondary variables were also measured todescribe the acceleration of the head. These two are the peak accelerationand the mean acceleration (plateau acceleration). The peak accelerationcorresponds to the overshoot seen in the beginning of the acceleration phaseand the plateau is the mean value after the peak. Figure 4.3 illustrates thevariables.

A repeated measure ANOVA was performed, using the primary variableTr, comprising of two level: stabilized and not stabilized. The ANOVA wasperformed at a significance level of α = 0.05 and observations was testedfor sphericity with Mauchly’s test. When the assumption of sphericity wasviolated this was corrected using Greenhouse-Geisser correction. The effectsize is described with η2partial and post hoc test performed with Tukeys’ HSD-test. See subsection 4.2.2 for details on statistics.[25, 26, 27, 28, 29]

19

(a) Device mounted (b) Stabilizer and head

(c) Stabilizing device

Figure 4.2: The stabilizing device used in study III. The device is mountedon the frame of the chair on the left side (right side as seen by the subject).It can be adjusted in order to fit tightly to the patients cheek

20

16800 17000 17200

-0,1

0,0

0,1

Volta

ge[V

]

Sample

Rise time

Plateau mean

Peak

Figure 4.3: Conceptual picture properties measured in study III. Rise timeis the time between the two small vertical lines. Peak acceleration is the scalarvalue at the peak and the plateau mean is the mean of every motion at theplateau. The signal has an offset from zero

21

22

5

Results

Results from the studies are presented individually under separate sections.Results from the ANOVAs are presented directly in the chapter and all posthoc tests are presented in appendices.

5.1 Study I

The results from the measurements can be seen in Figure 5.1 - 5.4, whichare further described as mean value and variance in Table 5.1. The chair didnot reach the desired acceleration but had a downward offset. Below followsthe ANOVA of this data.

5.1.1 Translational Motion ANOVA

The result from the ANOVA can be seen in Table 5.2 and the post hoc testsis in Appendix B.

The ANOVA showed interaction effects within weight*direction*acceleration(F(3.297, 62.640)=32.6 p<.001). The true acceleration (the one measured)was significant different for all levels of desired acceleration (F(3, 57)=2585*E2p<.001) and the post hoc tests give significant differences between all com-binations of acceleration (p<.001), see Figure 5.5 and Table B.1

The Effect of Weight

There is a significant difference within weights (F(1.395, 26.511)=220,95p<.001). The post hoc tests show that there is no significant difference be-tween empty and adult (p=.991) but a significant difference between emptyand child and between child and adult (both p<.001). See Figure 5.6 andTable B.2 for diagrams and post hoc results.

23

0 5 10 15 2042

43

44

45

46

47

Acce

leratio

n[cm/s^2

]

Sample

Translational50 cm/s^2Left

Empty Child Adult

(a)

0 5 10 15 20

89

90

91

92

93

94

95

96

Acce

leratio

n[cm/s^2

]

Sample

Translational100 cm/s^2Left

Empty Child Adult

(b)

0 5 10 15 20133

134

135

136

137

138

139

140

141

142

143

144

145

146

Acce

leratio

n[cm/s^2

]

Sample

Translational150 cm/s^2Left

Empty Child Adult

(c)

0 5 10 15 20183184185186187188189190191192193194195196197198199200201

Acce

leratio

n[cm/s^2

]

Sample

Translational200 cm/s^2Left

Empty Child Adult

(d)

Figure 5.1: Translational left acceleration

0 5 10 15 20

42

44

46

Acce

leratio

n[cm/s^2

]

Sample

Translational50 cm/s^2Right

Empty Child Adult

(a)

0 5 10 15 20

90

92

94

96

Acce

leratio

n[cm/s^2

]

Sample

Translational100 cm/s^2Right

Empty Child Adult

(b)

0 5 10 15 20

136

138

140

142

144

146

Acce

leratio

ncm

/s^2

Sample

Translational150 cm/s^2Right

Empty Child Adult

(c)

0 5 10 15 20180

182

184

186

188

190

192

194

196

198

200

202

Acce

leratio

n[cm/s^2

]

Sample

Translational200 cm/s^2Right

Empty Child Adult

(d)

Figure 5.2: Translational right acceleration

24

0 5 10 15 20

36

38

40

42

44

Acce

lera

tion

[deg

/s]

Sample

Rotational50 deg/s^2CW

Empty Child Adult

(a)

0 5 10 15 20

70

72

74

76

78

Acce

lera

tion

[deg

/s]

Sample

Rotational100 deg/s^2CW

Empty Child Adult

(b)

0 5 10 15 2098

100

102

104

106

108

110

112

114

116

118

120

Acce

lera

tion

[deg

/s]

Sample

Rotational150 deg/s^2CW

Empty Child Adult

(c)

0 5 10 15 20

126

128

130

132

134

136

138

140

142

144

146

148

150

152

154

Acce

lera

tion

[deg

/s]

Sample

Rotational200 deg/s^2CW

Empty Child Adult

(d)

Figure 5.3: Rotationall CW acceleration

0 5 10 15 2036

38

40

42

Acce

lera

tion

[deg

/s]

Sample

Rotational50 deg/s^2CCW

Empty Child Adult

(a)

0 5 10 15 2068

70

72

74

76

78

Acce

lera

tion

[deg

/s]

Sample

Rotational100 deg/s^2CCW

Empty Child Adult

(b)

0 5 10 15 2096

98

100

102

104

106

108

110

112

114

116

118

Acce

lera

tion

[deg

/s]

Sample

Rotational150 deg/s^2CCW

Empty Child Adult

(c)

0 5 10 15 20

124

126128

130

132

134

136138

140

142

144

146148

150

152

Acce

lera

tion

[deg

/s]

Sample

Rotational200 deg/s^2CCW

Empty Child Adult

(d)

Figure 5.4: Rotational CCW acceleration

25

Table 5.1: Mean acceleration and variance of measured motions

Motion Mean acceleration σ2

Empty Child Adult Empty Child Adult

Translational.Left.50 44.19 44.24 45.42 1.57 0.22 0.50Translational.Left.100 92.66 93.41 93.45 2.94 0.53 0.08Translational.Left.150 140.58 144.56 141.70 7.97 0.88 0.17Translational.Left.200 188.86 196.85 195.83 7.77 3.78 0.81

Translational.Right.50 44.55 44.70 43.89 0.40 0.71 0.06Translational.Right.100 93.05 93.87 90.59 2.35 0.16 0.59Translational.Right.150 141.06 141.41 138.08 3.21 0.32 0.71Translational.Right.200 187.54 194.75 183.23 7.86 3.30 1.37

Rotational.CW.50 39.57 39.73 39.39 2.78 1.18 1.12Rotational.CW.100 73.46 73.56 74.45 5.65 4.42 4.05Rotational.CW.150 106.18 106.83 108.79 25.48 29.87 30.54Rotational.CW.200 140.00 138.67 137.75 55.87 54.88 54.69

Rotational.CCW.50 38.74 38.89 38.67 1.31 1.16 1.32Rotational.CCW.100 73.05 73.03 73.51 3.57 4.16 4.91Rotational.CCW.150 104.86 105.94 105.85 21.32 23.67 23.08Rotational.CCW.200 136.30 134.57 137.76 45.58 51.61 56.72

26

Table 5.2: ANOVA of translational motions.

Effect SS DF MS F p η2partial

Weight 784 1.395 392 221.0 <.001 0.921Error 67 26.511 2Direction 504 1 504 114.9 <.001 0.858Error 83 19 4Acceleration 1432763 2.368 477588 258484.8 <.001 1.000Error 105 44.997 2Wei.*Dir. 606 1.354 303 189.6 <.001 0.909Error 61 25.727 2Wei.*Acc. 795 2.931 132 69.8 <.001 0.786Error 216 55.680 2Dir.*Acc. 498 1.985 166 85.5 <.001 0.818Error 111 37.722 2Wei.*Dir.*Acc. 393 3.297 65 32.6 <.001 0.632Error 229 62.640 2

Figure 5.5: Translational ANOVA, acceleration.

27

Figure 5.6: Translational ANOVA, weights.

Weight and acceleration have interaction effects (F(2.931, 55.680)=69.798p<.001) and the interaction increases with higher accelerations, see Fig-ure 5.7. The post hoc test support this; there is no significant differencebetween weights at low acceleration (50 cm/s2) but there is significant dif-ferences at higher accelerations. The post hoc test is seen in Table B.3.

The Effect of Direction

There is a significant difference between left and right motions (F(1, 19)=114.88p<.001), see Figure 5.8. The mean acceleration is higher for ”downhill” mo-tion (see section 3.1).

Direction and acceleration have interaction effects (F(1.985, 37.722)=85.461p<.001); see Figure 5.9. Post hoc test gives that there is no difference be-tween directions at 50 cm/s2 (p=.999) and 100 cm/s2 (p=.167) but a sig-nificant difference at 150 and 200 cm/s2 (both p<.001). See Table B.4 posthoc test.

The Effect of Direction and Weight

Interaction effects is found between direction and weight (F(1.354, 25.727)=189.56p<.001), see Figure 5.10. Post hoc test gives no significant difference betweendirection with an empty chair (p=.998) and a significant difference betweendirection with load child and adult (p<.001 for both). The difference betweendirection increases with higher load, see Table B.5 for post hoc test.

28

Figure 5.7: Translational ANOVA, weights*acceleration.

Figure 5.8: Translational ANOVA, direction.

29

Figure 5.9: Translational ANOVA, direction*acceleration.

Figure 5.10: Translational ANOVA, weight*direction.

30

Table 5.3: ANOVA of rotational motions.

Effect SS DF MS F p η2partial

Weight 234 1.097 117 10.118 .004 0.372Error 439 20.873 12Direction 64 1 64 8.085 .010 0.903Error 439 19 8Acceleration 650886 1.797 216962 1722.483 <.001 0.990Error 7180 34,137 126Wei.*Dir. 94 1.072 47 3.952 .058 0.123Error 451 20.359 12Wei.*Acc. 436 1.203 73 7.382 .009 0.424Error 1121 22.864 10Dir.*Acc. 386 1.141 129 12.276 .001 0.517Error 598 21.642 10Wei.*Dir.*Acc. 507 1.209 85 8.598 .005 0.466Error 1120 22.974 10

5.1.2 Rotational Motion ANOVA

The ANOVA can be seen in Table 5.3 and the post hoc tests in Appendix C.

The ANOVA showed interaction effects within weight*direction*acceleration(F(1.209, 22.974)=8.598 p=.005), see Figure 5.11. The true accelerationswere significant different (F(1.797, 34.137)=1722.5 p<.001) and the post hoctest says there is significant difference between all combinations of accelera-tion (p<.001 for all), see Table C.1.

The Effect of Weight

There is a significant difference between weights (F(1.099, 20.874)=10.118p<.001), see Figure 5.12, and the post hoc test says there is a significantdifference between the weights empty and child (p=.012) and between childand adult (p<.001). The weights empty and adult is significant (p=.379),see Table C.2 for post hoc test results.

Weight and acceleration show interaction effects (F(1.203, 22.864)=7.382p=.009), see Figure 5.13. Post hoc test (Table C.3) gives significant differencebetween weight only at an acceleration of 100 deg/s2 where child load issignificantly different from empty and adult (p<.001 for both). Otherwiseno significant differences is found.

31

Figure 5.11: Rotational ANOVA, acceleration.

Figure 5.12: Rotational ANOVA, weights.

32

Figure 5.13: Rotational ANOVA, weights*acceleration.

The Effect of Direction

CW and CCW motions are significant different (F(1, 19)=8.0854 p=.010), seeFigure 5.14. There are interaction effects between direction and accelerations(F(1.141, 21.672)=12276 p=.001), see Figure 5.15.

Post hoc test gives no significant differences between directions at 50 and150 deg/s2 acceleration (p=.876 and p=.092) and that there is a significantdifference between direction at an acceleration of 100 and 200 deg/s2 respec-tively (p=.012 and p=.001), see Table C.4.

The Effect of Weight and Direction

The ANOVA shows no interaction effects between direction and weight (F(1.072,20.359)=3.952 p=.058), see Figure 5.16. Since the observation shows no in-teraction effects, no post hoc tests are made.

5.2 Study II

The within set ANOVA, seen in Table 5.4, shows that there is significantdifferences within the set (F(4.067, 77.281)=105.67 p<.001), i.e. significantdifference between repetitions in a drive file, see Figure 5.17. Post hoc testcan be seen in Table D.1.

33

Figure 5.14: Rotational ANOVA, direction.

Figure 5.15: Rotational ANOVA, direction*acceleration.

34

Figure 5.16: Rotational ANOVA, weight*direction.

Figure 5.17: Within set ANOVA.

35

Table 5.4: ANOVA of within set repetitions.

Effect SS DF MS F p η2partial

Repetition 178.8 4.067447 19.9 105.667 <.001 0.848Error 32.1 77.28149 0.2

Table 5.5: ANOVA of between set.

Effect SS DF MS F p η2partial

Repetition 32 19 2 1.191 .297 0.284Error 81 57 1

The between set ANOVA, seen in Table 5.5, gives that there is no signif-icant difference between sets (F(19, 57)=1.1907 p=.297), i.e. no significantdifference between repetition of the same drive file, see Figure 5.18.

5.3 Study III

The effect of stabilization is clear (F(1, 19)=363 p<.001), see Figure 5.19 andTable 5.6 for the ANOVA. The mean rise time sink with just over 0.05 s from0.162 to 0.108 s and the post hoc test shows significant differences betweensubject when the head is not stabilized. With stabilization, no significantdifference can be seen between subjects, see Table E.1.

Table 5.6: ANOVA of head stabilization.

Effect SS DF MS F p η2partial

Subject 0.008343 2.755115 0.002086 4.3482 .010 0.186Error 32.1 52.34718 0.000480Stabilization 0.146882 1 0.146882 363.721 <.001 0,950Error 0.007673 19 0.000404Sub*Stab 0.018257 2.968330 0.004564 10.5914 <.001 0.358Error 0.032751 56.39826 0.000431

36

Figure 5.18: Between set ANOVA.

Figure 5.19: Rise time for subjects with and without stabilizing device

37

Table 5.7: Rise time, peak acceleration and plateau mean for head accelera-tion on subjects. Plateau value for subject two could not be measured. Valuesfor the chair is also included in column ”Stabilized”. Rise time has unit s,all other units are in cm/s2

Subject Not stabilized StabilizedRise time Peak Plateau Rise time Peak Plateau

Subject 1 0.157 472.2 166.8 0.113 324.3 120.6Subject 2 0.138 469.3 - 0.114 277.5 158.2Subject 3 0.186 582.5 265.7 0.105 312.7 125.7Subject 4 0.167 517.3 131.8 0.105 274.6 117.8Subject 5 0.160 494.1 144.8 0.101 276.6 120.2Mean 0.162 507.1 177.3 0.108 290.2 126.9

Chair 0.089 185.2 135.7

5.3.1 Acceleration of the Head

When the head is not stabilized the peak and plateau value reaches highervalues than the chair. The values are although all lower when the head isstabilized. All data can be read in Table 5.7.

38

6

Discussion

The chapter starts with a discussion of the results and how they could beinterpreted. This is followed by a discussion on statistical hypothesis testingand the chapter ends with suggestions for future work.

6.1 General Discussion

In this section the results from the studies are discussed in the context ofused method and what the aims of the thesis were. Briefly, the result maybe summarized as to: reject H01 and H02; weight and inclination impactthe performance; and the head can be stabilized.

6.1.1 Study I

Accuracy is the capability of the measurement system to quantify the truevalue. Accuracy can thus be described as a measure of bias, where an idealmeasurement system should have no bias. The precision is instead the degreeto which repeated measurements under unchanged conditions show the sameresults, in other words reproducibility and repeatability. The precision canbe described as a measure of how spread the values are (variance).

Table 5.1 shows that we have a bias; the true acceleration is lower thanthe desired one. The table also shows that the variance (measure of precision)vary with acceleration and to some degree also with weight. The precision in-creases with higher acceleration for both translational and rotational motionbut much more for rotational motions (Figure 5.5 and 5.11).

Figure 5.7 shows that the bias depends on weight when the motion istranslational. When the motion is rotational, the interaction between accel-eration and weight is significant (Figure 5.13) and the post hoc test shows

39

difference between weights only at an acceleration of 100 deg/s2 where childis significantly different from empty and adult. Over all it looks like weightdoes not interact with acceleration and since only one value is different thereis a risk that this is a measurement error.

The correlation between weight and acceleration is not linear for trans-lational motions. A reasonable assumption would have been that if weightand acceleration interacted then the acceleration should have decreased withincreasing weight since larger load requires more effect from the motors.However, this is not the case. As seen in Figure 5.6 and 5.7 the impact isnon-linear between load and acceleration. This result indicates that not onlythe effect of the motors causes variation but there is a possibility that thecause of it is within the software (control algorithms).

Over all, the results support a rejection of hypothesis one, H01, i.e. weightimpacts the accuracy and precision.

There is a difference between right and left direction of motion (Figure 5.8and 5.14). The difference is seen for both types of motions and a probablecause is the inclination found in section 3.1. The acceleration is higher for”downhill” motions (left and CW) and lower for uphill motions for bothtranslational and rotational motions. The post hoc tests add that the differ-ence between direction occurs at higher accelerations; perhaps an indicationthat the motors are too weak for the ”upper range” of acceleration. Theeffect size for direction is for rotational motion η2partial=0.903 and for trans-lational η2partial=0.858; which means when only using one motor (rotationalmotion) the direction (hence inclination) is a larger factor influencing thedifference.

The effect size and the post hoc test together indicate that the inclinationis an error that the drive line can not compensate for and that the motorsor control algorithm is not optimized in that area. Another problem withthe inclination is that the motion pattern is changed. The intention with thechair is to create motion which is perpendicular to the direction of gravity(translational motion) to predominantly stimulate Utricle. If the rail has aninclination there will be a vertical component in the motion which is not theintention.

6.1.2 Study II

Bias is not always a problem, if the precision is good the bias can be com-pensated for but if the precision is bad it is more difficult. For translationalmotion the bias is easy to compensate simply by specify a higher accelera-tion in the drive file. However, the same compensation is not easy to accessfor rotational motions where a cyclic behavior is found, The cyclic behavior

40

leads to large variance in output (Table 5.1). Large variance is undesirableand study II support the observation that there is a large within set variance.On the positive side, the between-set-repeatability is good; we have a largevariance but at least the variance does not vary between sets. The largevariance for rotational motion seen in study I can be explained by the cyclicbehavior. The variance for every repetition is not calculated but Figure 5.17gives a hint that it is low.

One solution to the cyclic behavior is to just have one repetition perset/drive file, but since it is desirable to run several repetitions without ini-tiating the chair between every run, the behavior is problematic.

6.1.3 Study III

In the last study the head was successfully stabilized. The variance in headstabilization in response to head acceleration was large, both within andbetween subjects. The head stabilization device significantly reduced thisvariance in repeated measures within subject and also between subjects (Ta-ble E.1).

The rise time correspond to the first reaction to acceleration but is onlyone variable in the measurement of stability. It is not easy to compare theacceleration curve between subject and chair since the chair is a mechani-cal system with high repeatability while the behavior of the human head issubjected to high inter- and intra-subject variability.

The vestibular system sends input to stabilize the head in the accelerationphase but the same happens for the vision and the somatosensory system.This means that it is possible to cancel out certain input to study each systemseparately. If the light is turned of and the head is stabilized, it is possibleto study passive elicitation of the VO; if the light is off and the head is notstabilized it is possible to study the interaction between VO and somatosen-sory system and so on. If increasing the precision in the secondary variablesit should be possible to more accurately study different input contributionto head stabilization.

Another possibility is to measure the head roll (rotation around axis ofsight) instead of translational motion. The SNR was low and possible rota-tional signals could not be distinguished from the noise. A rotational motionwill have a translational component and this is seen as a undershoot in thetranslational acceleration. A large undershoot leads to a longer rise timewhich means it is possible to measure the roll indirectly in the rise time oftranslational acceleration.

The results support a rejection of hypothesis two, H02.

41

6.2 Level of Significance

When performing hypothesis testing statistics there is one thing that alwaysshall be kept in mind: the meaning of p-value, for example what p<0.05actually means. Given that H0 is true, it gives the probability of the observeddata set. It does not say to ”what degree” the hypothesis is true or not, itcould rather be roughly described that it provides information how good adata set fits within a hypothesis.

In medicine it is common standard to use a significance level of α=0.05when including patients or other living things and hence the level of signifi-cance for study III was easy to decide.

Study I and II, on the other hand, is pure technical evaluations and alower α could have been used. The risk by using a lower α is that we mightrisk to miss differences that actually exist, in other word accepting a null-hypothesis that is wrong, i.e. do a type II-error. When using the higher αthe situation is the opposite; the risk is bigger to reject a true null-hypothesisby finding differences that does not exist, i.e. do a type I-error.

In study I and II an evaluation that aims to test performance and to findartifact was done. It was seen as less crucial to do a type I-error and find asmany potential problems as possible and try to filter among these and seewhat is most important and possible to fix.

6.3 Limitations and Suggestions

The first thing that should be fixed is the inclination of the rails. This shouldsolve much of the direction difference and most important it will create asolely horizontal motion. This is, probably, an easy and crucial fix for theMSP.

The cyclic behavior for rotational motion is perhaps the most disturbingerror. The problem lies probably within the software when translating drivefile in to some actuation protocol. The problem could be solved by designingdrive files with one or a few repetitions or by modifying the code translatinginto protocols. The cause could be within the motors as well but the motorseffect is oversized to handle accelerations in the range of this test.

Another problem is the creation of drive files with multiple repetitions;the system complains that the chair is out of range even if one is manuallycalibrating it not to be (see section 3.4). Solutions to the problem couldbe to enable inactivation of the safety-range-calculations or simply make thecalculations more accurate then they are today.

The stabilization study revealed that it is possible to stabilize the head

42

and to measure the motion of the head. The MSP together with the stabi-lizing device enable new studies of the different systems contributing to headstability. The stabilizing device should probably work for rotational motionas well and most certainly for translational motion in the other directionthan tested in study III.

43

44

7

Conclusions

The aim with the thesis is to evaluate the performance, i.e. accuracy andprecision, of a newly built motion simulation platform. The thesis testedtwo hypothesis: whether weight impacts the acceleration and if the headof a patient could be stabilized. The thesis ended up with the followingconclusions.

• The acceleration does not reach the desired value; BIRGIT does notshow high accuracy.

• The precision, measure of variance, varies with desired acceleration andmotion type.

– The variance for translational motion increases with increased ac-celeration.

– The variance for rotational motion is high. The acceleration fol-lows a cyclic behavior and the difference between repetitions withina set is high but variance between sets is low.

• Weight impacts the acceleration but not in the same way for rotationaland translational motions.

• The rails have an inclination of 0,5 deg and the acceleration is higherin ”downhill” direction.

• There are problematic constraints when drive file parameters are spec-ified.

• The head can be mechanically stabilized.

45

46

8

Acknowledgements

First of all I would like to thank all the staff at the Bernadotte laborato-ries/Karolinska Institutet and especially assoc. prof. Tony Pansell who madethis thesis possible and for his support. I really enjoyed my stay at thelaboratory!

I would also like to thank assoc. prof. Roberto Bolzani at the University ofBologna for helping me with signal processing and my friend Johan Suvanenwith whom I have discussed the thesis a lot. A big thank is sent to thosewho participated in the study as test persons.

At last, my supervisor at KTH, Chiara Giordano, shall have a specialthank as well.

Simon LundbergStockholm, Sweden

May, 2014

47

48

Bibliography

[1] Werner Kahle. Color Atlas/Text of Human Anatomy, volume 3. NervousSystem and Sensory Organs. Georg Thieme Verlag KG, 4th edition,1993.

[2] Kenneth J. Ciuffreda and Barry Tannen. Eye Movement Basic for theClinician. Mosby-Year Book, Inc., 1995.

[3] Alan Desmond. Vestibular Function: Clinical and Practice Management.Georg Thieme Verlag KG, 2nd edition, 2011.

[4] GM Halmagyi, T Haslwanter, IS Courthoys, RA Yavor, and MJ Todd.Three-dimensional vector analysis of the human vestibuloocular reflexin response to high-acceleration head rotations. ii. responses in subjectswith unilateral vestibular loss and selective semicircular canal occlusion.Journal of neurophysiology, 76(6):4021–4030, December 1996.

[5] Tony Pansell. Ocular counter-rolling during head tilt. PhD thesis,Karolinska Institutet, 2003.

[6] R. John Leigh and David S. Zee. The Neurology of Eye Movements.Oxford University Press, Inc., 4th edition, 2006.

[7] Schwartz Gary S. The Eye Exam: a Complete Guide. SLACK Incorpo-rated, 2006.

[8] Halmagyi GM and Curthoys IS. A clinical sign of canal paresis. ArchNeurol., 45(7), July 1988.

[9] F Wuyts. Principle of the head impulse (thrust) test or halmagyi headthrust test (hhtt). B-ENT, 8(4):23–25, 2008.

[10] Newman-Toker David E. Neuro-vestibular examination syllabus.Spencer S. Eccles Health Science Library, University of Utah, 2010.

49

[11] Stefano Ramat, David S. Zee, and Lloyd B. Minor. Translationalvestibulo-ocular reflex evoked by a ”head heave” stimulus. The vestibu-lar labyrinth in health and disease, 942, 2001.

[12] Lee Goldman and Andrew I. Schafer. Goldman’s Cecil Medicin. Saun-ders, 24 edition, 2011.

[13] J P Guyot and G Psillas. Test-retest reliability of vestibular autorota-tion testing in healthy subjects. Otolaryngol Head and Neck Surgery,117(6):704–7, 1997.

[14] T P Hirvonen, H Aalto, I Pyykko, and M Juhola. Comparison of twoautorotation tests. Journal of Vestibular Research, 9(2):119–25, 1999.

[15] P J Blatt, Schuberts M C, K E Roach, and R J Tusa. The reliability ofthe vestibular autorotation test (vat) in patients with dizziness. Journalof neurologic physical therapy, 32(2):70–9, Jun 2008.

[16] Barbara F. Lujan and Ronald J. White. Barany chair. Technical report,Science Teachers’ Association of Ontario, 2000.

[17] Leen Maes, Ingeborg Dhooge, Eddy De Vel, D’haenens, Annelies Bock-stael, Hannah Keppler, Birgit Philips, Freya Swinnen, and Bart M.Vinck. Normative data and test-retest reliability of the sinusoidal har-monic acceleration test, pseudorandom rotation test and velocity steptes. Journal of Vestibular Research, 18(4):197–208, February 2008.

[18] Joel A. Goebel, Permsarp Isipradit, and Jason M. Hanson. Manualrotational testing of the vestibulo-ocular reflex. The Laryngoscope,110(4):517–535, April 2000.

[19] National Space Biomedical Research Institute. Context spe-cific adaption of gravity dependent vestibular reflex responses.http://mvl.mit.edu/Research/Neurovestibular/Pages/project1.html,1998.

[20] J. Dits, M. M. J. Houben, and J. van der Steen. Three dimensionalvestibular ocular reflex testing using a six degrees of freedom motionplatform. Journal of Visualized Experiments, e4144(75):10.3791/4144,May 2013.

[21] Fredrik Rubensson and Bernt Lindholm. Construction of device forneurological research. Master’s thesis, Royal Institute of Technology,KTH, Stockholm, 2006.

50

[22] Patrik Osterlind. Design av styrsystem till rotationsstol och utvecklingav numerisk modell av otolitmembran. Master’s thesis, Royal Instituteof Technology, KTH, Stockholm, 2007.

[23] Jose Luis Guinon, Emma Ortega, Jose Garcıa-Anton, and ValentınPerez-Herranz. Moving average and savitzki-golay smoothing filter us-ing mathcad. In International Conference on Engineering Education -ICEE 2007, 2007.

[24] Alan V Oppenheim and Alan S Willsky. Signals and Systems. Prentice-Hall, Inc., 2nd edition, 1997.

[25] D. R. Cox. Principles of Statistical Inference. Cambridge UniversityPress, 2006.

[26] Richard G. Lomax and Debbie L. Hahs-Vaughn. Statistical concepts: asecond course. Routledge, fourth edition, June 2013.

[27] E. R. Girden. ANOVA: repeated measures. SAGE Publications, Inc,Newbury Park, California, 1992.

[28] Charles A. Pierce, Richard A. Block, and Herman Aguinis. Caution-ary note on reporting eta-squared values from multifactor anova design.Educational and Psychological Measurements, 64(6):916–924, December2004.

[29] Yosef Hochberg and Ajit C. Tamhane. Multiple Comparison Procedures.John Wiley and Sons, Inc, 1987.

[30] Hygh D. Young, Roger A. Freedman, and A. Lewis Ford. Sears andZemansky’s University Physics: with Modern Physics. Addison-Wesley,12th edition, March 2007.

[31] John J. Swearingen. Determination of centres of gravity of man. Tech-nical report, Civil Aeromedical Reserach Institute, Federal AviationAgency, August 1962.

[32] John J. Swearingen and Joseph W. Young. Determination of centresof gravity of children, sitting and standing. Technical report, CivilAeromedical Reserach Institute, Federal Aviation Agency, August 1965.

[33] A. Fredriks, S. van Buuren, W. J. M. van Heel, R. Dijkman-Neerincx,S.P. Verloove-Vanhorick, and J. M. Wit. Nationwide age references for

51

sitting height, leg length, and sitting height/height ratio, and their diag-nostic value for disproportionate growth disorder. Diseases in Childhood,90(8):807–812, 2005.

52

Appendix A

The Dummies

The most important characteristic of the dummy is its weight and for rota-tional motion the distribution of weight must be taken in to account. Twodummies are built, one modeling an adult (80 kg) and one modeling a child(40 kg, which is approximately a 12 year old).

To model the weight distribution the center of mass is used. This virtualpoint is the mean position the mass distributed in an object. In mechanicalcalculations this point is used as the point on which forces act on and is thecalled center of gravity (CG). When actuating rotational movement only onemotor is in action since the point of rotation is the upper link point and ithence stands still. The torque and effect vary with the distance between cen-ter or rotation and where the mass is located. This means it is of importanceto place the mass correctly.[30]

The rotation shall be around around the nasio (the nasal root) of thepatient. Swearingen et al. [31, 32] have measured the CG of humans. Whenadults sits down with their hands in lap, their CG is located 21,5 cm abovethe seat botton and for children this number is 20 cm. Fredriks et al. [33]measured the sitting height at different ages. The sitting height (lengthbetween the vortex of the head and seat bottom) is 79 cm in 12 year oldchildren and 95 cm for adults. The distance between the vortex and thenasio is approximated to be 10 cm for adult and 7 cm for children. Thisgives us the distance between the upper link point and the CG to be: 79-7-21=51 cm for children and 95-10-21.5=63.5 cm for adults

The dummies were built with four sandbags (2*25 kg and 2*15 kg). Thebags were distributed over the chair and restrained with straps as can be seenin Figure A.1. The CG of the dummies were adjusted to coincide with thecalculated ones and the best fit was 51 cm and 63 cm the point of rotationfor child and adult dummy, calculations is seen in Equation A.1-A.2. Thedistance lx corresponds to the distance between center of rotation and the

53

(a) Adult set up, 80 kg (b) Child set up, 40 kg

Figure A.1: Weight distribution. Green dot=Center of rotation, Dark bluedot=25kg, Light blue dot=15kg

bags middle point.

CGAdult =msmall(l1 + l2) +mlarge(l3 + l4)

mtotal

=15(43 + 98) + 25(55 + 65)

80= 63

(A.1)

CGChild =msmalll5 +mlargel6

mtotal

=15 ∗ 40 + 25 ∗ 58

40= 51 (A.2)

54

Appendix B

Post hoc tests, study Itranslational

Below is all results from the post-hoc tests computed in study I for transla-tional movement. The significance level is α = 0, 05 ∀ tests.

Table B.1: Translational ANOVA post-hoc test of accelerations. Tukey’sHSD-test. Probabilities for post-hoc tests error: Within MS=1,8476,DF=57,000, Mean value in table

Acceleration 50 100 150 20044.42 92.84 141.23 191.18

50100 <.001150 <.001 <.001200 <.001 <.001 <.001

Table B.2: Translational ANOVA post-hoc test of weights. Tukey’s HSD-test. Probabilities for post-hoc tests error: Within MS=1,7730, DF=38,000,Mean value in table

Weight Empty Child Adult116.50 119.22 116.52

EmptyChild <.001Adult .991 <.001

55

Table

B.3:

Tran

slational

AN

OV

Apost-hoc

testof

weight*acceleration

.T

ukey’s

HS

D-test.

Probabilities

forpost-hoc

testserror:

Within

MS

=1,8979,

DF

=114,000,

Mean

value

intable

Weigh

t*Acceleration

E*50

E*100

E*150

E*200

C*50

C*100

C*150

C*200

A*50

A*100

A*150

A*200

44.1592.86

140.82188.20

44.4793.64

142.99195.80

44.659202

139.89189.53

Em

pty

*50E

mp

ty*100

<.001

Em

pty

*150<

.001<

.001E

mp

ty*200

<.001

<.001

<.001

Ch

ild*50

.996<

.001<

.001<

.001C

hild

*100<

.001.324

<.001

<.001

<.001

Ch

ild*150

<.001

<.001

<.001

<.001

<.001

<.001

Ch

ild*200

<.001

<.001

<.001

<.001

<.001

<.001

<.001

Ad

ult*50

.891<

.001<

.001<

.001.999

<.001

<.001

<.001

Ad

ult*100

<.001

.234<

.001<

.001<

.001<

.001<

.001<

.001<

.001A

du

lt*150<

.001<

.001.120

<.001

<.001

<.001

<.001

<.001

<.001

<.001

Ad

ult*200

<.001

<.001

<.001

.002<

.001<

.001<

.001<

.001<

.001<

.001<

.001

56

Table

B.4:

Tra

nsl

atio

nal

AN

OV

Apo

st-h

octe

stof

dire

ctio

n*a

ccel

erat

ion

.T

uke

y’s

HS

D-t

est.

Pro

babi

liti

esfo

rpo

st-h

octe

sts

erro

r:W

ithi

nM

S=

1,94

26,

DF

=57

,000

,M

ean

valu

ein

tabl

e

Dir

ecti

on*A

ccel

erat

ion

L*5

0L

*100

L*1

50L

*200

R*5

0R

*100

R*1

50R

*200

44.4

793

.17

142.

2819

3.85

44.3

892

.51

140.

1818

8.50

Lef

t*50

Lef

t*10

0<

.001

Lef

t*15

0<

.001

<.0

01L

eft*

200

<.0

01<

.001

<.0

01R

ight*

50.9

99<

.001

<.0

01<

.001

Rig

ht*

100

<.0

01.1

67<

.001

<.0

01<

.001

Rig

ht*

150

<.0

01<

.001

<.0

01<

.001

<.0

01<

.001

Rig

ht*

200

<.0

01<

.001

<.0

01<

.001

<.0

01<

.001

<.0

01

57

Table B.5: Translational ANOVA post-hoc test of weight*direction. Tukey’sHSD-test. Probabilities for post-hoc tests error: Within MS=1,5974,DF=38,000, Mean value in table

Weight*Direction E*L E*R C*L C*R A*L A*R116.46 116.55 119.77 118.68 119.10 113.95

Empty*LeftEmpty*Right .998Child*Left <.001 <.001Child*Right <.001 <.001 <.001Adult*Left <.001 <.001 .022 .314Adult*Right <.001 <.001 <.001 <.001 <.001

58

Appendix C

Post hoc tests, study Irotational

Below is all results from the post-hoc tests computed in study I for rotationalmovement and for the additional rotational study. The significance level isα = 0, 05 ∀ tests.

Table C.1: Rotational ANOVA post-hoc test of accelerations. Tukey’s HSD-test. Probabilities for post-hoc tests error: Within MS=1,8476, DF=57,000,Mean value in table

Acceleration 50 100 150 20039.16 72.13 106.41 137.51

50100 <.001150 <.001 <.001200 <.001 <.001 <.001

59

Table C.2: Rotational ANOVA post-hoc test of weights. Tukey’s HSD-test.Probabilities for post-hoc tests error: Within MS=11,565, DF=38,000, Meanvalue in table

Weight Empty Child Adult89.02 87.86 89.53

EmptyChild .012Adult .379 <.001

60

Table

C.3:

Rot

atio

nal

AN

OV

Apo

st-h

octe

stof

wei

ght*

acce

lera

tion

.T

uke

y’s

HS

D-t

est.

Pro

babi

liti

esfo

rpo

st-h

octe

sts

erro

r:W

ithi

nM

S=

1,89

79,

DF

=11

4,00

0,M

ean

valu

ein

tabl

e

Wei

ght*

Acc

eler

atio

nE

*50

E*1

00E

*150

E*2

00C

*50

C*1

00C

*150

C*2

00A

*50

A*1

00A

*150

A*2

0039

.15

73.2

510

5.52

138.

1539

.31

69.1

310

6.38

136.

6239

.03

74.0

210

7.32

137.

76

Em

pty

*50

Em

pty

*100

<.0

01E

mp

ty*1

50<

.001

<.0

01E

mp

ty*2

00<

.001

<.0

01<

.001

Ch

ild

*50

.999

<.0

01<

.001

<.0

01C

hil

d*1

00<

.001

<.0

01<

.001

<.0

01<

.001

Ch

ild

*150

<.0

01<

.001

.985

<.0

01<

.001

<.0

01C

hil

d*2

00<

.001

<.0

01<

.001

.567

<.0

01<

.001

<.0

01A

du

lt*5

0.9

99<

.001

<.0

01<

.001

.999

<.0

01<

.001

<.0

01A

du

lt*1

00<

.001

.995

<.0

01<

.001

<.0

01<

.001

<.0

01<

.001

<.0

01A

du

lt*1

50<

.001

<.0

01.3

11<

.001

<.0

01<

.001

.974

<.0

01<

.001

<.0

01A

du

lt*2

00<

.001

<.0

01<

.001

.999

<.0

01<

.001

<.0

01.8

97<

.001

<.0

01<

.001

61

Table

C.4:

Rotation

alA

NO

VA

post-hoctest

ofdirection

*acceleration.

Tu

key’sH

SD

-test.P

robabilitiesfor

post-hoctests

error:W

ithinM

S=

10,485,D

F=

57,000,M

eanvalu

ein

table

Direction

*Acceleration

CW

*50C

W*100

CW

*150C

W*200

CC

W*50

CC

W*100

CC

W*150

CC

W*200

39.5671.05

107.26138.80

38.7673.22

105.55136.21

CW

*50C

W*100

<.001

CW

*150<

.001<

.001C

W*200

<.001

<.001

<.001

CC

W*50

.876<

.001<

.001<

.001C

CW

*100<

.001.012

<.001

<.001

<.001

CC

W*150

<.001

<.001

.092<

.001<

.001<

.001C

CW

*200<

.001<

.001<

.001.001

<.001

<.001

<.001

62

Appendix D

Post hoc tests, study II

On the following page is the post hoc test computed in study II. The signif-icance level is α = 0, 05.

63

Table

D.1:

Within

setA

NO

VA

,post-hoc

testof

repetitions.

Tu

key’sH

SD

-test.P

robabilitiesfor

post-hoctests

error:W

ithinM

S=

11,565,D

F=

38,000,M

eanvalu

ein

table

Rep

etition1

23

45

67

89

1040.74

39.4738.56

37.9637.78

37.6638.04

38.3739.21

39.64

Rep

1R

ep2

<.001

Rep

3<

.001<

.001R

ep4

<.001

<.001

<.001

Rep

5<

.001<

.001<

.001.952

Rep

6<

.001<

.001<

.001.486

.998R

ep7

<.001

<.001

.006.999

.666.149

Rep

8<

.001<

.001.933

.076.001

<.001

.318R

ep9

<.001

.664<

.001<

.001<

.001<

.001<

.001<

.001R

ep10

<.001

.970<

.001<

.001<

.001<

.001<

.001<

.001.056

64

Appendix E

Post hoc tests, study III

On the following page is the post hoc test computed in study III. The sig-nificance level is α = 0, 05.

65

Table

E.1:

Stabilization

study

AN

OV

A,

post-hoctest

ofsu

bject*stabilization.

Tu

key’sH

SD

-test.P

robabilitiesfor

post-hoctests

error:W

ithinM

S=

0,00043,D

F=

76,000,M

eanvalu

ein

table

Sub

ject*Stab

ilization1*N

1*Y2*N

2*Y3*N

3*Y4*N

4*Y5*N

5*Y0.157

0.1130.138

0.1140.186

0.1050.167

0.1050.160

0.101

1*No

1*Yes

<.001

2*No

.128.011

2*Yes

<.001

.999.014

3*No

.001<

.001<

.001<

.0013*Y

es<

.001.952

<.001

.934<

.0014*N

o.879

<.001

.001<

.001.125

<.001

4*Yes

<.001

.941<

.001.921

<.001

.999<

.0015*N

o.999

<.001

.035<

.001.007

<.001

.990<

.0015*Y

es<

.001.711

<.001

.666<

.001.999

<.001

.999<

.001

66