Analysis of Mouse Whisker Movement Synchronicity

IN

DEGREE PROJECT MEDICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2020 ,

Analysis of Mouse Whisker Movement Synchronicity

JOHANNA ELISABETH DE VOOGEL

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ENGINEERING SCIENCES IN CHEMISTRY,

BIOTECHNOLOGY AND HEALTH

Analysis of Mouse Whisker Movement Synchronicity

JOHANNA ELISABETH DE VOOGEL

Master in Medical Engineering Date: March 10, 2020

Supervisor: Arvind Kumar & Gilad Silberberg Examiner: Madelen Fahlstedt

School of Engineering Sciences

in Chemistry, Biotechnology and Health

Swedish title: Analys av musens synkroniserade rörelse av

morrhåren

iii

Abstract

In active sensing, sensory organs are actively controlled by the motor system to optimize stimuli acquisition. Coupling behaviour of animal models with electrophysiological signals could give us important insights in the workings of this active sensing in health and disease. In this work, a contribution is made towards this aim. A modern machine learning method, called DeepLab- Cut, is evaluated and optimized for the use of tracking whiskers from videos of head-fixed, free whisking mice from two groups. Novel to the approach is the longer observation of the mice, without markers or trimming. This track- ing is then used to investigate the natural statistics of whisking in healthy and Parkinsons Disease (PD) modeled animals. Two groups of mice (6 healthy versus 4 PD) were filmed for 5 minutes on 1-3 days per animal during head- fixed free air whisking, resulting in 21 filmed days in total. 6 Whiskers were tracked with 4 points per whisker using DeepLabCut and as few as 105 man- ually labeled images. A tracking accuracy of <3 pixels training error and <

15 pixels test error was achieved. Qualitatively, tracking errors occurred both along and perpendicular to the whisker, but the perpendicular error reduced after 30 Hz low-pass filtering. The observed whisking was non-stationary and predominantly slow (<10 Hz). Whisking was also synchronized in time, as the cross-spectra were significantly higher than the random shuffled cross-values (p _KS <0.05), but this synchronicity was lost on the right side in lesion ani- mals with whisking frequencies between 15 and 25 Hz. In this PD-modeled group, a difference between left and right side co-synchronization was found (J SD=0.067, p _KS <0.05) for higher frequencies (15-25 Hz). In total, con- trol animals where whisking 34.1% of the time versus 35.6% for the lesion animals. Upon examining these active bouts, it was found that lesion animals have relatively more long bouts of co-activation than healthy controls across all subgroups (p _KS <0.001). Furthermore, in lesion animals the activity dis- tributions are laterally different (LL vs RR, J SD=0.178, p _KS <0.001), but this cannot be concluded for healthy controls. Thus, it can be concluded that DeepLabCut is a promising method to speed up whisker tracking. Future im- provements could be made in sharpness and contrast of the tips of the whiskers;

incorporating information on the relation of whisker position in time; extend-

ing the tracking to 3D; and evaluating tracking by means of biomechanical

models. These improvements will enable observation in a more realistic en-

vironment, and for a longer time. Finally, coupling of whisker position data

to electrophysiological recordings will hopefully result in new insights in the

relation between sensorimotor behaviour and global brain dynamics.

v

Sammanfattning

Vid ’active sensing’ styrs sinnesorgan aktivt av motorsystemet för att optimera insamling av stimuli. Koppling av djurmodeller med elektrofysiologiska sig- naler skulle kunna ge oss viktig insikt i funktionen av denna ’active sensing’ i hälsa och sjukdom. I detta arbete görs ett bidrag till detta mål. En modern ma- skininlärningsmetod, kallad DeepLabCut, utvärderades och optimerades för användning av spårning av morrhår från videor där möss i två grupper hade huvudet fixerat men morrhåren kunde rör sig fritt. Nytt för denna metod är den längre observationen av mössen utan märkning eller hårklippning. Denna spårning användes sedan för att undersöka den naturliga statistiken över morr- hårens rörelse i kontrollgruppen och gruppen med Parkinsons sjukdom (PD).

Två grupper av möss (6 kontroll mot 4 PD) filmades i 5 minuter på 1-3 dagar per djur där huvudet var fixerat men morrhåren kunde röra sig fritt, vilket re- sulterade i totalt 21 filmade dagar. Sex morrhår spårades med 4 punkter per hår med hjälp av DeepLabCut med så få som 105 manuellt märkta bilder. En spår- ningsnoggrannhet med träningsfel på <3 pixlar och testfel på <15 pixlar upp- nåddes. Kvalitativt inträffade spårningsfel både längs med och vinkelrätt mot morrhåret, men det vinkelräta felet minskade efter 30 Hz lågpassfiltrering. Den observerade rörelsen av morrhåret var föränderlig och övervägande långsam (<10 Hz). Rörelsen av morrhåren synkroniserades också i tid, eftersom kors- spektra var signifikant högre än de blandade korsvärdena (p _KS <0, 05), men denna synkronitet förlorades på höger sida i PD djur med frekvenser mellan 15 och 25 Hz. I denna PD grupp hittades en skillnad mellan vänster och höger sida i samsynkronisering (J SD=0, 067, p _KS <0, 05) för högre frekvenser (15- 25 Hz). Kontrollgruppen rörde på morrhåren i totalt 34,1 % av tiden mot 35,6

% för PD gruppen. Inom denna aktiva period visade det sig att PD djur hade relativt mer långa perioder av samaktivering än kontrolldjur i alla delgrupper (p _KS <0.001). Dessutom var aktivitetsfördelning laterallt olika i PD gruppen (LL vs RR, J SD=0, 178, p _KS <0, 001), men detta kunde inte påvisas i kon- trollgruppen. Således kan man dra slutsatsen att DeepLabCut är en lovande metod för att påskynda spårning av rörelse av morrhår. Framtida förbättringar skulle kunna göras med förbättrad skärpa och kontrast av morrhårens ändar;

innehålla information om morrhårens postion relativt till varandra över tid; ut-

vidga spårningen till 3D; och utvärdera spårning med hjälp av biomekaniska

modeller. Dessa förbättringar möjliggör observation i en mer realistisk miljö

och under en längre tid. Slutligen kommer koppling av morrhårspositionens

data till elektrofysiologiska inspelningar förhoppningsvis att resultera i ny in-

sikt i förhållandet mellan sensorimotoriskt beteende och global hjärndynamik.

Contents

Abstract iii

Sammanfattning v

List of Figures ix

List of Tables xiii

Acknowledgements xiv

1 Introduction 1

1.1 Problem description and problem statement . . . . 2

1.2 The importance of studying whisking behaviour . . . . 3

1.3 Target readers and report structure . . . . 3

2 Methods 4 2.1 Whisking data . . . . 4

Experimental model and subject details . . . . 4

Video specifications . . . . 5

2.2 Whisker tracking with DeepLabCut . . . . 5

2.3 Analysis of synchronicity . . . . 7

Angle computation . . . . 7

Spectral cross-correlation . . . . 8

Activity . . . . 9

Activity duration comparison . . . 10

3 Results 11 3.1 Whisker tracking evaluation . . . 11

3.2 Healthy behaviour . . . 15

Frequency analysis . . . 16

vi

CONTENTS vii

Spectral cross-products . . . 16

3.3 Comparison of lesion animals versus controls . . . 20

Activity and co-activity . . . 21

4 Discussion 24 4.1 Interpretation of the work . . . 24

4.2 Conclusions . . . 26

4.3 Future work . . . 26

Bibliography 28 A State of the Art 31 A.1 Introduction to the background . . . 31

A.2 Whisker anatomy . . . 31

Location . . . 31

Musculature . . . 32

Geometry . . . 34

Afference . . . 34

Efference . . . 34

A.3 Whisker function . . . 35

A.4 Behaviour and dynamics terminology . . . 36

Whisking cycle . . . 36

Whisking modes . . . 37

Angle definition . . . 37

Asymmetry in whisking . . . 38

Biomechanical models . . . 39

A.5 Tracking methods comparison . . . 41

Current whisker tracking methods . . . 41

Common challenges and method design choices . . . 41

The argument for DeepLabCut . . . 42

List of Figures

2.1 Set-up of the experiment. Mice were head-fixed with a clip in a tube, and filmed from a distance of approximately 30 cm on a white background surface. . . . . 6 2.2 DeepLabCut workflow adapted from Mathis et al. [13] . . . . 7 2.3 Schematic of angle computation, with manual labels (+), fit-

ted whiskers (black lines), base vectors of the left side (blue arrows) and right side (red arrows). The resulting azimuthal angle is displayed in green. . . . 8 2.4 Activity segmentation process. From top to bottom: com-

puted azimuthal angle (θ) time series after filtering (black);

normalized signal (black) and threshold (blue) at 1.2; thresh- olded signal convolved with an exponential kernel (black) and threshold (red) at 0.03; original time series (black) with the resulting segmented active periods (green). . . . . 9 3.1 Loss-convergence of the two training networks. Finding the

best network is an optimization problem, and corresponds to minimizing the loss-function. . . 12 3.2 The 3 test images from the control animals. Manual labels

are marked with a cross (+), predicted labels with a dot (.) in the following colors: L1 (blue), L2 (cyan), L3 (green), R1 (yellow), R2 (orange) and R3 (red). Eyes and nose are marked in dark blue. . . . 13 3.3 The 3 test images from the lesion animals. Manual labels are

marked with a cross (+), predicted labels with a dot (.) in the following colors: L1 (blue), L2 (cyan), L3 (green), R1 (yellow), R2 (orange) and R3 (red). Eyes and nose are marked in dark blue. . . 14

ix

x LIST OF FIGURES

3.4 Position coordinates in pixels for 4 points of whisker L2 (left- central) in a healthy animal. P1-P4 are the traces for the pre- dicted proximal to distal whisker points. The x-coordinates (light blue), y-coordinates(dark blue), and segmented activa- tion (green) are shown. . . 15 3.5 Computed azimuthal angle traces (black) for all 6 tracked whiskers

in a healthy animal, low-pass filtered at 30 Hz. Active whisk- ing bouts are highlighted in green. Only a short subset (15 seconds) of the entire video is shown. . . . 16 3.6 Angle series displayed on top of the power spectral density

for frequencies <30 Hz. Frequency resolution of 1 Hz, time windows of 1 second without overlap. The same 15 seconds for the healthy animal are used. . . 17 3.7 Cross-product for a healthy animal (blue) and a lesion animal

(red), computed by summing the products of corresponding time-windows of the spectrograms for every frequency band.

The diagonal is the auto-cross-product. The mean of the shuf- fled cross-spectra ± 1 standard deviation are shaded in blue (control) and red (PD). . . 17 3.8 Control group average cross-products for frequencies <30 Hz,

grouped in seven subgroups. Shaded bands are the mean shuf- fled values ± 1 standard deviation. . . 19 3.9 Lesion group average cross-products for frequencies <30 Hz,

grouped in seven subgroups. Shaded bands are the mean shuf- fled values ± 1 standard deviation. . . 19 3.10 Right-side (RR: R1R2, R1R3, R2R3) average cross-products

for frequencies <30 Hz for control and lesion animals. Shaded bands are the mean shuffled values ± 1 standard deviation over the lesion and control group. Individual animals are shown with dots. . . 20 3.11 Cumulative distribution (cdf, shaded) and probability distri-

bution (pdf, line) of active whisking bouts duration. Individ- ual control animals (blue, dashed) are compared with individ- ual lesion animals (red, dashed), as well as the average of the groups (solid lines). . . . 22 3.12 Cumulative distribution (cdf) and probability distribution (pdf)

of active whisking bouts duration for control and lesion ani-

mals, grouped in 7 subgroups. . . 23

LIST OF FIGURES xi

A.1 Principles for mystacial pad architecture. Reprinted from Brecht, Preilowski, and Merzenich [1] with permission . . . 31 A.2 Whisker anatomy . . . 33 A.3 Average whisker shaft angle during the triphasic whisking cy-

cle. Reprinted from Hill et al. [9]. . . . 36

A.4 Angle definitions . . . 37

List of Tables

2.1 Data acquisition overview. Filmed days are marked with an X. 5 2.2 Grouping for activity duration. L=left, R=right, X=cross-wise,

1=caudal, 2=central, 3=rostral . . . 10 3.1 Average Euclidean error between manual labels and predicted

labels in pixels, computed over n frames. . . 12 A.1 Comparison of whisker tracking methods. Adapted from Perkon

et al. [24] and Nashaat et al. [25]. . . 40

xiii

Acknowledgements

For this project, I would like to thank the following people. To start, of course, I want to thank my supervisor Arvind Kumar for his helpful advice and ex- planations on the computational side of the project, and Gilad Silberberg for the physiological interpretations. Also, the people at Karolinska Institutet and SciLifeLab have been of important value. I want to thank Roberto specifically, for filming the mice and helping with the data acquisition, and Johanna, for the pointers in the literature. Next, I would like to thank Athira for helping me to get started with DeepLabCut and to explain the cluster computing, and the people at HPC2N for helping me with any further questions I had. Further- more, the group supervision meetings with Xiaogai, Geraldine, Mathilde and Nicole have helped me in the process of writing my master thesis. Lastly, but not least, I want to thank my parents, my sisters, and Sven, for their emotional support.

xiv

Chapter 1 Introduction

How do we understand the world around us? What active strategies do we use to gather information, which our brain can then interpret? In active sensing, sensory organs are actively controlled by the motor system to optimize stimuli acquisition. Therefore we need to understand the sensorimotor system. A first step to this understanding, is to find how the sensory organs are operated and to find the differences in behaviour of sensorimotor systems in health and disease.

For example, one could study Parkinson’s disease (PD). PD is a nervous system disorder with motor symptoms such as tremor, rigidity and bradykine- sia. The prevalence of PD is estimated to be 100 to 200 per 100,000 people worldwide [1], and 196 per 100,000 inhabitants of Stockholm [2]. The aging population will only lead to more patients with PD.

Coupling behaviour of animal models with electrophysiological signals could give us important insights in the workings of active sensing [3]. Mice actively use whiskers, tactile hair organs, to gather information about their surroundings.

Mice have several rows of whiskers which differ in shape and length ac- cording to their position [4]. Each whisker is wrapped by an intrinsic mus- cle, and the mystacial pad on which the whiskers are situated is moved by a set of extrinsic muscles [5]. Whisking emerges through triphasic rhythmic protraction (forward motion) and retraction (backward motion) [6]. Whiskers have no spindles, but instead mechanoreceptors on follicles could transduce information about position [7]. One hypothesis is that the angular position of the whisker encodes the contact with a surface. Frequency and amplitude of the whisking differentiate several modes of whisking. In foveal whisking, the hairs move in low amplitude cycles with a frequency of 15-25 Hz. Ex-

1

2 CHAPTER 1. INTRODUCTION

ploratory whisking is characterized by large amplitude, low frequency (5-15 Hz) sweeps [8]. Terms that describe asymmetry of whisking are for example

‘contact induced asymmetry’, ‘head-turning asymmetry’, ‘rapid cessation of protraction’, and ‘angular separation’ [9].

1.1 Problem description and problem state- ment

The aim of this study was twofold: to track whiskers from videos in head- fixed, free whisking mice from two groups, without markers or trimming; and to investigate the natural statistics of whisking in healthy and PD animals.

First, the whiskers need to be tracked in enough detail. Thus far, this tracking has been problematic for multiple reasons. One reason is that, since whisking is fast, high-speed video is required, which has only become available recently.

Another reason is that whiskers are very thin and thus require high resolution images. These two combined create the further problem of big size of data, and could explain why previous work has focused on short whisking videos.

Lighting conditions can cause unwanted shadows on the background which are hard to distinguish from the whiskers. Furthermore, the different rows of whiskers often cross, overlap or move out of focus. Attempts to overcome these problems are trimming or labeling the whiskers, but they are sub-optimal as they might change the behaviour of mice and movement of all whiskers together [10]. However, the biggest limitation in tracking whiskers has been that manually tracking whiskers is very time-consuming. Recently, methods have been developed to make tracking (semi)-automatic, with varying results [11, 12]. DeepLabCut is such a machine learning method to track animal behaviour [13, 14]. The novelty of this work lies in the application of this method specifically to mouse whisker tracking, in more natural circumstances like untrimmed free-whisking for a longer time.

Furthermore, because tracking of the whiskers has been problematic, a thorough review of whisking behaviour in freely moving mice during a longer period is still lacking. To the best of my knowledge, no research is published on the difference in whisking behaviour in PD.

The goal of this thesis was to track and analyze untrimmed and unmarked

whisking, using only videos. Thus, an existing tracking method called DeepLab-

Cut was evaluated for the specific use of mouse whisker tracking. A second

goal was to measure synchronicity between whiskers and compare this syn-

chronicity in whisking behaviour between healthy and PD-modeled mice.

CHAPTER 1. INTRODUCTION 3

1.2 The importance of studying whisking be- haviour

Studying the sensorimotor system helps to understand how we gather infor- mation about the world, and more specifically how we recognize objects. Fur- thermore, we need insights in behaviour under PD to understand the disease and thus to help to improve the quality of life of patients.

The rationale for studying the mouse whisker behaviour is as follows. The whisker system is a well-known sensorimotor system, that has been studied ex- tensively. In particular, the output behaviour, whisking, is directly observable.

However, a detailed description of whisking behaviour is still missing (owing to the previously described problems). Just like humans use their fingers to probe an object, mice can actively use their whiskers to locate themselves, de- tect objects, and differentiate surfaces. Similar mechanisms might be in place during active surface exploration [15], and tactile object localization [16]. It is therefore expected that the whisking system is comparable to active sens- ing in humans and thus makes a suitable model. Furthermore, the human PD condition can be modeled by inducing a lesion in mice, through injection of 6-hydroxydopamine (6OHDA) resulting in dopamine-depletion [17]. This allows for certain experiments that would have been impossible in humans.

Moreover, longer observations in an experimental setting are possible.

1.3 Target readers and report structure

This report was written for a master’s degree of 30 ECTS in medical engi- neering. The purpose of this work was to improve tracking of whiskers us- ing videos, and to characterize whisking behaviour statistics in healthy and diseased mice. Thus, it is largely intended for medical engineering students.

However, it could also be useful to researchers in the field of neuroscience, machine learning, computational ethology, and of course to anyone studying whisking behaviour directly.

The report is structured as follows. By now, some key definitions and back-

grounds are introduced. A more extensive review of the State-of-the-Art can

be found in Appendix A. In the next section, the methods for filming, tracking

and analysing whisking are explained. The results are split into a first part

for the accuracy of the tracking, a second part for healthy whisking behaviour

and a third part for the comparison of whisking behaviour for healthy versus

lesioned mice.

Chapter 2 Methods

2.1 Whisking data

Experimental model and subject details

For the PD model, the procedure of Ketzef et al. [17] was used: All experi- ments were performed according to the guidelines of the Stockholm municipal committee for animal experiments under an ethical permit to Gilad Silberberg (N12/15). A D1-Cre (EY262 line) mouse line was crossed with the Channel- rhodopsin (ChR2)-YFP reporter mouse line (Ai32, the Jackson laboratory) to induce expression of ChR2 in either dMSNs or iMSNs. Female mice were housed under a 12 hr light-dark cycle with food and water ad libitum. All experiments were carried out during the light phase.

Mice (4 females, 8-10 weeks of age) were anesthetized with isoflurane and mounted in a stereotaxic frame (David Kopf Instruments, Tujunga, California).

Mice received one unilateral injection of 1 ml of 6OHDA-HCl (3.75 mg/ml dissolved in 0.02% ascorbic acid) into the medial forebrain bundle, accord- ing to the following coordinates (in mm) [18]: antero-posterior -1.2, medio- lateral +1.2 and dorso-ventral -4.8. Only 6OHDA-injected mice that showed rotational behavior [19] were used in the experiments. Unlesioned mice (6 females, 8-10 weeks of age) were used as a healthy control group.

The 6 healthy and 4 lesion animals were filmed over several days. An overview of the filmed days can be found in Table 2.1. A clip was mounted to each animal for head-fixing, but no whiskers were trimmed or marked.

4

CHAPTER 2. METHODS 5

Table 2.1: Data acquisition overview. Filmed days are marked with an X.

Animal / Day 1 2 3 4 5 Lesion

1 x x

2 x x x

3 x x x

4 x x

Control

1 x x

2 x

3 x x

4 x

5 x x x

6 x x

Video specifications

All mice were free-whisking, meaning that the whiskers were not obstructed by any object. To approximate natural behaviour, the animals were not stim- ulated to whisk, as is commonly achieved by for example air-puffs or object localization tasks in return for rewards. The camera was placed overhead, and whisker movements in the horizontal plane were recorded with a resolution of 896 x 432 pixels at a distance of approximately 30 cm. The set-up is shown in Figure 2.1. For each day, 5 videos of 1-minute duration were filmed consecu- tively, with 50, 200 and 400 fr/sec. For the final analysis, 400 fr/sec was used because whiskers were least blurred wit this frame rate. After experimenting with different lighting conditions, white light was chosen over infrared light as it resulted in the best contrast and visibility of the individual top-row whiskers.

2.2 Whisker tracking with DeepLabCut

For the tracking of the whiskers, DeepLabCut/2.0.4-Python-3.6.6 was used [14] [13], [20]. This is an algorithm which uses a pretrained pose-estimation network, to train a deep neural network to track features, using only limited training data.

The pipeline of DeepLabCut, adapted to this research, is shown in Figure

2.2. To create a training and test set, 6 whiskers were labeled once manually

6 CHAPTER 2. METHODS

Figure 2.1: Set-up of the experiment. Mice were head-fixed with a clip in a tube, and filmed from a distance of approximately 30 cm on a white back- ground surface.

with 4 points on each whisker. The tracked whiskers were chosen by their size,

horizontal orientation and visibility, and attempted to be the same across dif-

ferent animals and different filming days. Furthermore, the nose and eyes were

labeled for the angle computation (see next section). 5 Dissimilar frames were

selected with K-means from the first 6 seconds of every video and consecu-

tively labeled for each day for each mouse, leading to a total of 50 manually

labeled frames for the lesion animals, and 55 manually labeled frames for the

control animals. An example of the labeled frames is shown in Figure 2.3. The

network was trained with ResNet50 for 200000 iterations, using 47 and 52 of

the labeled images for lesion animals and control animals, respectively. The

average Euclidean error was computed by comparing the predicted labels with

the manual labels for 3 test images. Finally, all videos were analysed using the

newly trained network. Consequently, the output x- and y-coordinates for each

label were merged across the five 1-minute videos per day per animal, to be

used for further analysis.

CHAPTER 2. METHODS 7

dlc-models Directory

created

File created

Process Create project

Extract frames

Label frames (GUI)

videos labeled-data

training-datasets

config.yaml

Create training datasets

Train network

Evaluate network

No

Yes

Good Results

Yes

No

Need more training data?

Analyze video stop

Analyze video Extract outlier frames

Refine labels (GUI) Merge datasets

evaluation-results

HPC2N computing

cluster Windows

GPU

Figure 2.2: DeepLabCut workflow adapted from Mathis et al. [13]

2.3 Analysis of synchronicity

Angle computation

For each timepoint, the angle of the whisker was computed as follows. First,

a third-degree polynomial was fitted with least-squares through the 4 points

of each whisker. As Knutsen, Biess, and Ahissar [21] showed, the whisker

can be approximated with a quadratic curve, and thus moves predominantly

in a flat (horizontal) plane. Furthermore, the shape of the whisker remained

the same during head-fixed, free air whisking. Therefore, the azimuthal angle

was defined as the angle between the tangent of this fit at the first whisker point

8 CHAPTER 2. METHODS

𝜃𝑅 𝜃𝐿

L1

L2 R3 L3

R2

R1 Right eye Left eye

Nose

L2-p1

L2-p2

L2-p3 L2-p4

+ +

L1-p2

L1-p3

L1-p4

L3-p1 L3-p2

L3-p3 L3-p4 R1-p3

R1-p2 R1-p4

R2-p2 R2-p1

R2-p3 R2-p4

R3-p1 R3-p3

R3-p4

R3-p2

Figure 2.3: Schematic of angle computation, with manual labels (+), fitted whiskers (black lines), base vectors of the left side (blue arrows) and right side (red arrows). The resulting azimuthal angle is displayed in green.

and the line between the nose and respective eye. This was done to account for mystacial pad and nose movements. Right and left side whiskers were mirrored so that a protraction of whiskers always corresponds to a positive direction. Figure 2.3 illustrates this process. Following Towal and Hartmann [22], to reduce tracking noise, the estimated angle time series were low-pass filtered (f _{cut−of f} = 30Hz) with a Butterworth filter.

Spectral cross-correlation

The resulting time-angle series were examined in the frequency domain by comparing their spectrograms. As a measure of synchronized whisking, cross- correlograms were computed for each pair of spectrograms within animals for each day. However, due to the sparsity of the signal (mice were not whisk- ing for most of the filmed time), the power of most frequencies was low. To overcome this problem, cross-spectrograms were computed as follows: For each time window the synchronicity correlation (spectral cross-product) was defined as:

1 N

P N

k=1 S ₁ (t _k , f ) ·S ₂ (t _k , f )

Where N is the number of windows in the total signal, S1 and S2 are the

power spectra, and f are the frequencies. To see if the resulting correlation

was higher at the same time window than at any other time (in other words,

if the whiskers are whisking with the same frequency at the same time, or at

any time), (n=7600) random permutations were carried out and the shuffled

CHAPTER 2. METHODS 9

correlations were computed and averaged. Thus, in these shuffles, different time windows were multiplied and summed, again for each frequency.

Figure 2.4: Activity segmentation process. From top to bottom: computed az- imuthal angle (θ) time series after filtering (black); normalized signal (black) and threshold (blue) at 1.2; thresholded signal convolved with an exponential kernel (black) and threshold (red) at 0.03; original time series (black) with the resulting segmented active periods (green).

Activity

Active whisking periods were extracted according to the following steps. First, the time-series of the angles were normalized, and thresholded at 1.2 standard deviations of the mean. Then, the resulting signal was convolved with an ex- ponential kernel (with a smoothing window of 50 frames, and kernel size of 2000 frames or 5 seconds). Finally, the signal was segmented at a threshold of 0.03. The percentage of activity was computed as the number of frames la- beled as active in this way, divided by the total number of frames.The process is shown in Figure 2.4.

The signals of all animals and all days were binned in two groups, one

healthy group and one lesion group. To compare them, a two-sided Kolmogorov-

Smirnov two-sample test (KS) was performed for each pair of groups to deter-

mine if there was a significant difference.

10 CHAPTER 2. METHODS

Activity duration comparison

In order to study same-side and cross-side synchronicity, the durations of ac- tive whisking bouts were compared. For this, the co-activity of each pair of whiskers was defined as the collection of all bout lengths at which the pair was whisking at the same time. It was computed as the product of the activity of each pair. The pairs were then grouped according to Table 2.2.

Here, too KS-tests were performed. To measure the similarity between each pair between and within groups and subgroups, for all probability distri- butions, the Jensen-Shannon divergence (JSD) was computed as:

D _{J S} (P, Q) = ¹ ₂ D _KL (P, M ) + ¹ ₂ D _KL (Q, M ), where M = ¹ ₂ (P + Q)

and D KL is the Kullback-Leibler divergence (KL), computed as:

D _KL (P, Q) = ^P _x∈χ P (x) log ^{ P (x)} _Q(x) ^ . In contrast to KL, JSD is symmet- ric and for base 2 (which is used to compare two probability distributions), it ranges between 0 and 1.

Table 2.2: Grouping for activity duration. L=left, R=right, X=cross-wise, 1=caudal, 2=central, 3=rostral

Group Pairs

L L1L1,L2L2,L3L3

R R1R1,R2R2,R3R3

LL L1L2,L1L3,L2L3

RR R1R2,R1R3,R2R3

LR L1R1,L2R2,L3R3

XLR L1R2,L1R3,L2R3

XRL L3R1,L2R1,L3R2

Chapter 3 Results

As mentioned before, the whisking system is studied in a wide range of set- tings. Improvements in automatization of the tracking of individual whiskers would have a big impact in this research field, in the case of a higher accu- racy and efficiency of tracking. Using machine learning to track whiskers is expected to increase the accuracy compared to the current automatic tracking.

Furthermore, machine learning approaches obviate the need for markers, dyes and hair trimming.

It is important to study synchronicity of whisker movement, because this could give insights in whether whiskers are independently controlled, whether mice are lateralized (have a preferential side of whisking initiation), and whether whisker movements could influence each other. When coupled to electrophys- iological data, this information could help gain a deeper understanding of the sensorimotor system.

3.1 Whisker tracking evaluation

Finding the optimal weights for the neural network is an optimization algo- rithm. The objective is to minimize the tracking error, which means to mini- mize the loss. To evaluate whether the optimal weights for the network have been found, the loss-convergence can therefore be inspected. Figure 3.1 shows the loss-function, which indeed converges for both trained networks. The test- and training-error, computed as the average distance between the manual la- bels and the predicted labels of the lesion and control network are shown in Table 3.1. The labeling errors of the test images are shown in Figure 3.2 and Figure 3.3 for healthy controls and lesion animals respectively. The observed shifts of the predicted label along the whisker compared to the manual label

11

12 CHAPTER 3. RESULTS

will still lead the polynomial fit to follow the whisker shape and thus do not affect the angle computation. In contrast, errors perpendicular to the whisker result in misdetection and thus a deviation from the true whisking angle (for example in Figure 3.2, R3’s most distal point in the third image). One disad- vantage of DeepLabCut is that it traces frame-by-frame, so big jumps of a label between two consecutive frames can occur. This seemed to happen predom- inantly in the distal (p4) points. To account for this, the signal was low-pass filtered, but potentially some high-frequency whisking characteristics might have been lost here.

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018

0 20000 40000 60000 80000 100000 120000 140000 160000 180000 200000

Lo ss

Iteration

Loss-convergence

Lesion Control

Figure 3.1: Loss-convergence of the two training networks. Finding the best network is an optimization problem, and corresponds to minimizing the loss- function.

Table 3.1: Average Euclidean error between manual labels and predicted labels in pixels, computed over n frames.

Control Lesion

Train 2.24 (n=52 fr) 1.72 (n=47 fr)

Test 11.42 (n=3 fr) 14.64 (n=3 fr)

CHAPTER 3. RESULTS 13

Figure 3.2: The 3 test images from the control animals. Manual labels are marked with a cross (+), predicted labels with a dot (.) in the following colors:

L1 (blue), L2 (cyan), L3 (green), R1 (yellow), R2 (orange) and R3 (red). Eyes

and nose are marked in dark blue.

14 CHAPTER 3. RESULTS

Figure 3.3: The 3 test images from the lesion animals. Manual labels are marked with a cross (+), predicted labels with a dot (.) in the following colors:

L1 (blue), L2 (cyan), L3 (green), R1 (yellow), R2 (orange) and R3 (red). Eyes

and nose are marked in dark blue.

CHAPTER 3. RESULTS 15

3.2 Healthy behaviour

520

x 540

p1

600

x 700

p2

500

x 750

p3

0 50 100 150 200 250 300

time [s]

250 500

x 750

p4

160 180 y

100 200 300 y

100 200 300 y

0 250 y

Figure 3.4: Position coordinates in pixels for 4 points of whisker L2 (left- central) in a healthy animal. P1-P4 are the traces for the predicted proximal to distal whisker points. The x-coordinates (light blue), y-coordinates(dark blue), and segmented activation (green) are shown.

To inspect the tracking quality of the individual whisker points, the label coordinates were plotted. Figure 3.4 shows an example of the output label coordinates (in pixels) of four marked points on one left caudal whisker (L2) for a healthy mouse, with the active periods marked in green. Note that as the predicted points are more distal, (p1 closest to the shaft, p4 furthest away from the shaft), there are several large spike-like deviations which are likely to be tracking errors. Low-pass filtering to get rid of these deviations is therefore sensible.

The computed low-pass filtered angles for a subset of the same animal are

shown in Figure 3.5. Remarkably, the whiskers do not seem to be active at the

same time, for the same length. Naively, one would expect whiskers to whisk

at the same time. However, this could be partially caused by an activation

detection mismatch, instead of reflecting true activation difference.

16 CHAPTER 3. RESULTS

0

[ ] 100 L1

100 0

[ ] L2

0

[ ] 100 L3

0

[ ] 100 R1

0

[ ] 100 R2

126 128 130 132 134 136 138 140

time [s]

0

[ ] 100 R3

Figure 3.5: Computed azimuthal angle traces (black) for all 6 tracked whiskers in a healthy animal, low-pass filtered at 30 Hz. Active whisking bouts are highlighted in green. Only a short subset (15 seconds) of the entire video is shown.

Frequency analysis

It was observed that the angle time series were non-stationary. Therefore, the spectrograms were computed with a frequency resolution of 1 Hz and time windows of 1 second without overlap. An example of the previous signal can be seen in Figure 3.6. Since the signal was filtered, only the power spectral density of frequencies <30 Hz are shown. The frequency content indeed varies considerably over time, but the lower frequencies (0-10 Hz) have more power.

Spectral cross-products

In Figure 3.7, the cross-products averaged over all time windows of 1 second

are shown for one control animal and one lesion animal. A higher value cor-

responds to a higher synchronization of two whiskers averaged over all time-

windows for a specific frequency band of 1 Hz. The lower frequencies have

a higher spectral cross-product, meaning the whiskers are more synchronized

in the lower frequencies.

CHAPTER 3. RESULTS 17

0 25 freq L1

0 25 freq L2

0 25 freq L3

0 25 freq R1

0 25 freq R2

126 128 130 132 134 136 138 140

time [s]

0 25 freq R3

0

[ ] 100

0.0 0.5

100 0

[ ]

0.0 0.5

0

[ ] 100

0.0 0.5

0

[ ] 100

0.0 0.5

0

[ ] 100

0.0 0.5

0

[ ] 100

0.0 0.5

Figure 3.6: Angle series displayed on top of the power spectral density for frequencies <30 Hz. Frequency resolution of 1 Hz, time windows of 1 second without overlap. The same 15 seconds for the healthy animal are used.

Figure 3.7: Cross-product for a healthy animal (blue) and a lesion animal

(red), computed by summing the products of corresponding time-windows of

the spectrograms for every frequency band. The diagonal is the auto-cross-

product. The mean of the shuffled cross-spectra ± 1 standard deviation are

shaded in blue (control) and red (PD).

18 CHAPTER 3. RESULTS

Figure 3.8 shows the average cross-products for the subgroups for all pairs

of the healthy animals together. The shaded areas under the curves are the

mean of the permutations ± 1 standard deviation. Subgroups L and R theoret-

ically have a cross-product equal to or greater than the synchronization sub-

groups. When treating the cross-spectra as distributions, and performing the

KS test and computing JSD of these normalized distributions, the following

results are obtained: The shuffle is lower than the auto-cross-spectra for both

sides (p _KS <0.001). There is a difference between shuffle and cross-spectra in

all subgroups (p _KS <0.05), except for the cross-wise subgroup XLR. Thus, in

general, there is a certain synchronization in time in the whiskers. Although

left might seem more synchronized than right, no significant differences are

found within the control group between left and right or between unilateral and

bilateral whisking (p _KS >0.05). However, the JSD between L and R versus LL

and RR (0.270 vs 0.029) is remarkable.

CHAPTER 3. RESULTS 19

5 10 15 20 25 30

freq [Hz]

10 ⁶ 10 ⁵ 10 ⁴ 10 ³ 10 ² 10 ¹

average cross-product

control

L LL R RR LR XLR XRL

Figure 3.8: Control group average cross-products for frequencies <30 Hz, grouped in seven subgroups. Shaded bands are the mean shuffled values ± 1 standard deviation.

5 10 15 20 25 30

freq [Hz]

10 ⁶ 10 ⁵ 10 ⁴ 10 ³ 10 ² 10 ¹

average cross-product

lesion

L LL R RR LR XLR XRL

Figure 3.9: Lesion group average cross-products for frequencies <30 Hz,

grouped in seven subgroups. Shaded bands are the mean shuffled values ±

1 standard deviation.

20 CHAPTER 3. RESULTS

5 10 15 20 25 30

freq[Hz]

10 ⁶ 10 ⁵ 10 ⁴ 10 ³ 10 ² 10 ¹

average cross-product

RR

control lesion

control per mouse lesion per mouse

Figure 3.10: Right-side (RR: R1R2, R1R3, R2R3) average cross-products for frequencies <30 Hz for control and lesion animals. Shaded bands are the mean shuffled values ± 1 standard deviation over the lesion and control group. In- dividual animals are shown with dots.

3.3 Comparison of lesion animals versus con- trols

Figure 3.9 shows the average cross-products for the subgroups for all pairs of the lesion animals together. Again, KS-test and JSD are performed. KS-tests reject the null-hypothesis that the distributions are the same for all compar- isons of subgroups between control and lesion animals. Here, too, there is a difference between shuffle and cross-spectra in all subgroups (p _KS <0.05), but not in the right side synchronization (RR). Similar to the controls, no signifi- cant difference is detected between left and right side whiskers by the KS-test (p KS >0.05). When inspecting the lower frequencies (5-15 Hz) and higher frequencies (15-25 Hz) separately, still no difference is found in the lower fre- quencies but in the higher frequencies there is a difference between LL and RR (J SD=0.067, p KS <0.05).

No significant difference is detected when comparing between control and

lesion animals for all subgroups (p _KS >0.05). However, the JSD is bigger for

CHAPTER 3. RESULTS 21

the right side than for the left side, both in auto-cross-spectra (R=0.303 vs L=0.208), and synchronization spectra (RR=0.163vs LL=0.133). A difference is found in the lower frequencies (5-15 Hz) on both left side (L, J SD=0.143, p _KS <0.01), and right side (R, J SD=0.487, p _KS <0.05). For the higher fre- quencies (15-25 Hz), the difference is only there in L and RR (J SD=0.080 and J SD=0.112, p KS <0.05). To illustrate this difference between the two fre- quency groups, the right-side synchronization is shown in Figure 3.10. These results suggest that in lesion animals, the high frequency (15-25 Hz) synchro- nization is lost on the right side.

Activity and co-activity

After low-pass filtering (cut-off frequency = 30 Hz) and segmenting the active bouts, control animals were found to be actively whisking 34.1% of the time, and lesion animals 35.6% of the time.

The duration distributions are different between controls and lesion an- imals for all subgroups (p KS <0.001). The JSD between healthy and PD- modeled mice is bigger for the right side than for the left side (R,J SD=0.205;

L, J SD=0.181) and this relation is kept in the co-activation durations (RR, J SD=0.237; LL, J SD=0.178). In all cases, lesion animals have more in- stances of longer simultaneous active periods. This is illustrated in Figure 3.11.

Within group analysis is shown in Figure 3.12b. Within the control group, a small difference is observed between left (L) and right (R) active durations (L vs R, J SD=0.145, p KS <0.01). In the co-activation, this difference is not significant (LL vs RR, J SD=0.130, p KS >0.05). However, in the lesion group, the difference is bigger, both between left-right (L vs R, J SD=0.174, p _KS <0.001) and after co-activation of the sides (LL vs RR, J SD=0.178, p KS <0.001).

When comparing the bilateral whisking (LR) with both unilateral sides, the difference with the right side is bigger than with the left side, for both the healthy animals (LR vs LL, J SD=0.120, p _KS <0.01; LR vs RR, J SD=0.130, p KS <0.001) and the lesion animals (LR vs LL, J SD=0.132, p KS <0.01; LR vs RR, J SD=0.150, p KS <0.001).

Both in the control and lesion animals, there is significant difference be-

tween the activation distribution and the co-activation distributions, both uni-

laterally and bilaterally (p KS <0.001). This means that the whiskers are not

always active for the same time.

22 CHAPTER 3. RESULTS

10

10

10

10

duration [s]

10

10

2 × 10

3 × 10

4 × 10

6 × 10

pdf

LL, ks = 0.0

0.0 0.2 0.4 0.6 0.8 1.0

cdf

control merged lesion merged control per mouse lesion per mouse

(a) Unilateral left side co-activation (LL: L1L2, L1L3, L2L3).

10

10

10

10

duration [s]

10

10

2 × 10

3 × 10

4 × 10

6 × 10

pdf

RR, ks = 0.0

0.0 0.2 0.4 0.6 0.8 1.0

cdf

control merged lesion merged control per mouse lesion per mouse

(b) Unilateral right side co-activation (RR: R1R2, R1R3, R2R3).

10

10

10

10

duration [s]

10

10

2 × 10

3 × 10

4 × 10

6 × 10

pdf

LR, ks = 0.0

0.0 0.2 0.4 0.6 0.8 1.0

cdf

control merged lesion merged control per mouse lesion per mouse

(c) Bilateral co-activattion (LR: L1R1, L2R2, L3R3).

Figure 3.11: Cumulative distribution (cdf, shaded) and probability distribution

(pdf, line) of active whisking bouts duration. Individual control animals (blue,

dashed) are compared with individual lesion animals (red, dashed), as well as

the average of the groups (solid lines).

CHAPTER 3. RESULTS 23

10 ² 10 ¹ 10 ⁰ 10 ¹

duration [s]

10 ² 10 ¹ 10 ⁰

pdf

0.0 0.2 0.4 0.6 0.8 1.0

cdf

control

L LL R RR LR XLR XRL

(a) Control animals.

10 ² 10 ¹ 10 ⁰ 10 ¹

duration [s]

10 ² 10 ¹ 10 ⁰

pdf

0.0 0.2 0.4 0.6 0.8 1.0

cdf

lesion

L LL R RR LR XLR XRL

(b) Lesion animals.

Figure 3.12: Cumulative distribution (cdf) and probability distribution (pdf)

of active whisking bouts duration for control and lesion animals, grouped in 7

subgroups.

Chapter 4 Discussion

4.1 Interpretation of the work

To summarize, it was shown that tracking of mouse whiskers with an accu- racy of <3 pixels training error and < 15 pixels test error was achieved by us- ing DeepLabCut and labeling as few as 105 images for a total of 5 minutes of observed head-fixed whisking for 10 animals on 21 days. Qualitatively, track- ing errors occurred both along the whisker and perpendicular to the whisker, but the perpendicular error seemed reduced after 30 Hz low-pass filtering.

The observed whisking was non-stationary and predominantly slow (<10 Hz).

Whisking was also synchronized in time, as the cross-spectra were signifi- cantly higher than the random shuffled cross-values (p KS <0.05), but this syn- chronicity was lost on the right side in lesion animals with whisking frequen- cies between 15 and 25 Hz. In this PD-modeled group, a difference between left and right side co-synchronization was found (J SD=0.067, p _KS <0.05) for higher frequencies (15-25 Hz). In total, control animals where whisking 34.1% of the time versus 35.6% for the lesion animals. Upon examining these active bouts, it was found that lesion animals have relatively more long bouts of co-activation than healthy controls. This difference holds across all sub- groups (p _KS <0.001). Furthermore, in lesion animals the activity distributions are laterally different (LL vs RR, J SD=0.178, p KS <0.001), but this cannot be concluded for healthy controls. Next, the observations are discussed in more detail.

There was a difference between training and test error. However, only 3 im- ages were used to compute this error. To decrease the test error, more labeled images could be added. More consistency in labeling will probably also de- crease the test error. Furthermore, the video quality could be increased, more

24

CHAPTER 4. DISCUSSION 25

specifically whiskers moving out of focus should be addressed. For example, a camera with higher resolution at bigger distance could be used. Although already very low, more consistent labeling at equidistant points along whisker would likely reduce the training error, as the marked points closer to the tip seem to be falsely located upon visual inspection. It was already pointed out that reducing human labeling variability will lower the error by the makers of DeepLabCut [14]. The loss has converged, so more training-iterations would not help.

Furthermore, now the labeling was done by one person. Using a labeling average of multiple experts could improve the tracking and reduce the error due to inconsistently tracking whiskers. Now, the grouped analyses (left/right, caudal/central/rostral) could be contaminated since perhaps non-homologous whiskers were tracked bilaterally, or across different days, or between different animals.

Because the method to segment active whisking bouts is rather basic, it can introduce several unwanted effects. One problem is that the thresholds were chosen visually and are thus rather unspecific. Low amplitude whisking bouts might therefore have been wrongly excluded. Another example is that individual bouts might have been split into several segments. This can be ex- tended to the co-activity, where splitting of one whisker effects the co-activity duration distribution. A comparison between two other methods of activity segmentation (HMM and TICC) has been done before, but discussion of their advantages and limitations is beyond the scope of this work.

Compared to Wallach et al. [23], less active whisking was observed (they reported 76,4%). This could have influenced the dynamics of whisking, be- cause the animals were perhaps in a different activity state. The difference in activity can also be explained by the experimental paradigm, as Wallach et al.

[23] trimmed all but three whiskers, computed the angle relative to a baseline instead of to the head, and video acquisition was triggered manually, perhaps at more active episodes. Bout durations had a similar range, but were slightly shorter (median of around 0.1 s versus 0.412 s in [23].

While Towal and Hartmann [22] and Mitchinson et al. [24] found differ-

ences between the left and right side in synchronicity, this was not apparent in

these cross-spectra. Importantly, Towal and Hartmann [22] found that these

asymmetries were attached to head movements. The animals in this experi-

ment were head-fixed, which could explain the lack of tested difference. How-

ever, sliding movements of the nose to one direction were observed. These

could indicate attention shifts from the animals to one side, as a compensation

for head motion.

26 CHAPTER 4. DISCUSSION

Bermejo, Friedman, and Zeigler [25] and Hill et al. [6] reported that whisk- ing on the same side of the face was synchronous. This matches the found similarity in the cross-products between unilateral whiskers, but contradicts Sachdev, Sato, and Ebner [26].

Interestingly, the observed difference in exploratory (5-15 Hz) and foveal (15-25 Hz) whisking by Berg and Kleinfeld [8] also holds with regard to lateral synchronization of lesion animals. While not enough is known to explain this lack of synchronization in the right side, it must be noted that the PD-model is created by inducing a lesion in only one hemisphere, which could correspond to a change between left and right side whiskers. The difference in duration distribution between sides in the PD-model but not in the controls reinforces this.

4.2 Conclusions

Thus, it can be concluded that DeepLabCut is a promising method to speed up whisker tracking, but there are several drawbacks. One is that the tracking accuracy is dependent on the video quality, specifically sharpness and contrast of the tips of the whiskers. Furthermore, information on the relation of whisker position in time is not used and might thus introduce errors.

From this study, whisking seems to be slow (<10 Hz) and non-stationary.

A difference between left and right side whisking synchronization cannot be concluded based on the presented results. Activation distributions however, are different between PD-modeled mice and healthy controls, with the PD- modeled mice whisking for longer durations together. Furthermore, PD-modeled mice lose, according to this study, unilateral synchronicity on the right side, together with a sidedness in activation duration distributions.

4.3 Future work

The future improvements that Datta et al. [3] suggested working on in compu-

tational ethology are well applicable in the context of this research. Improve-

ments in the tracking can be made by extending the tracking to 3D. Here, I

would like to add that biomechanical models of whisker movement could eval-

uate automatic whisker tracking methods, along with the comparison with the

current golden standard, namely human labeling. Furthermore, these improve-

ments in tracking will then enable observation in a more realistic environment,

and enable observation for a longer time, so that behaviour in hierarchical

CHAPTER 4. DISCUSSION 27

states like sleep/activity and hunger/thirst can be researched.

Finally, another next step to take would be the coupling of whisker posi-

tion data to electrophysiological recordings. This will hopefully result in new

insights in the relation between sensorimotor behaviour, and global brain dy-

namics.

Bibliography

[1] Ole Bjørn Tysnes and Anette Storstein. “Epidemiology of Parkinson’s disease”. In: Journal of Neural Transmission 124.8 (Aug. 2017), pp. 901–

905.

[2] J. Lökk et al. “Drug and treatment costs in Parkinson’s disease patients in Sweden”. In: Acta Neurologica Scandinavica 125.2 (Feb. 2012), pp. 142–

147.

[3] Sandeep Robert Datta et al. “Computational Neuroethology: A Call to Action”. In: Neuron 104.1 (Oct. 2019), pp. 11–24.

[4] Michael Brecht, Bruno Preilowski, and Micháel M. Merzenich. “Func- tional architecture of the mystacial vibrissae”. In: Behavioural Brain Research 84.1-2 (Mar. 1997), pp. 81–97.

[5] J Dörfl. “The musculature of the mystacial vibrissae of the white mouse.”

In: Journal of anatomy 135.Pt 1 (Aug. 1982), pp. 147–54.

[6] Dan N. Hill et al. “Biomechanics of the vibrissa motor plant in rat:

Rhythmic whisking consists of triphasic neuromuscular activity”. In:

Journal of Neuroscience 28.13 (Mar. 2008), pp. 3438–3455.

[7] Lauren E. McElvain et al. “Circuits in the rodent brainstem that con- trol whisking in concert with other orofacial motor actions”. In: Neuro- science 368 (Jan. 2018), pp. 152–170.

[8] Rune W. Berg and David Kleinfeld. “Rhythmic whisking by rat: Re- traction as well as protraction of the vibrissae is under active muscular control”. In: Journal of Neurophysiology 89.1 (Jan. 2003), pp. 104–117.

[9] Ben Mitchinson et al. “Active vibrissal sensing in rodents and marsupi- als”. In: Philosophical Transactions of the Royal Society B: Biological Sciences 366.1581 (Nov. 2011), pp. 3037–3048.

28

BIBLIOGRAPHY 29

[10] Jason Wolfe et al. “Texture coding in the rat whisker system: Slip-stick versus differential resonance”. In: PLoS Biology 6.8 (Aug. 2008). Ed.

by Garrett B Stanley, pp. 1661–1677.

[11] Igor Perkon et al. “Unsupervised quantification of whisking and head movement in freely moving rodents”. In: Journal of Neurophysiology 105.4 (Apr. 2011), pp. 1950–1962.

[12] Mostafa A. Nashaat et al. “Pixying behavior: A versatile real-time and post hoc automated optical tracking method for freely moving and head fixed animals”. In: eNeuro 4.1 (Jan. 2017), ENEURO.0245–16.2017.

[13] Alexander Mathis et al. “DeepLabCut: markerless pose estimation of user-defined body parts with deep learning”. In: Nature Neuroscience 21.9 (2018), pp. 1281–1289.

[14] Tanmay Nath et al. “Using DeepLabCut for 3D markerless pose esti- mation across species and behaviors”. In: Nature Protocols 14.7 (July 2019), pp. 2152–2176.

[15] Mariama Dione and Johan Wessberg. “Human 8- To 10-Hz pulsatile motor output during active exploration of textured surfaces reflects the textures’ frictional properties”. In: Journal of Neurophysiology 122.3 (Sept. 2019), pp. 922–932.

[16] Guy Horev et al. “Motor-sensory convergence in object localization:

A comparative study in rats and humans”. In: Philosophical Transac- tions of the Royal Society B: Biological Sciences 366.1581 (Nov. 2011), pp. 3070–3076.

[17] M. Ketzef et al. “Dopamine Depletion Impairs Bilateral Sensory Pro- cessing in the Striatum in a Pathway-Dependent Manner”. In: Neuron 94.4 (2017), 855–865.e5.

[18] F. Paxinos and K.B.J Franklin. The Mouse Brain in Stereotaxic Coor- dinates. Academic Press, 2001.

[19] Emanuela Santini et al. “Critical Involvement of cAMP/DARPP-32 and Extracellular Signal-Regulated Protein Kinase Signaling in l-DOPA- Induced Dyskinesia”. In: Journal of Neuroscience 27.26 (2007), pp. 6995–

7005. eprint: https://www.jneurosci.org/content/27/

26/6995.full.pdf.

[20] Alexander Mathis and Richard Warren. “On the inference speed and

video-compression robustness of DeepLabCut”. In: bioRxiv (2018), p. 457242.

30 BIBLIOGRAPHY

[21] Per Magne Knutsen, Armin Biess, and Ehud Ahissar. “Vibrissal Kine- matics in 3D: Tight Coupling of Azimuth, Elevation, and Torsion across Different Whisking Modes”. In: Neuron 59.1 (July 2008), pp. 35–42.

[22] R. Blythe Towal and Mitra J. Hartmann. “Right-left asymmetries in the whisking behavior of rats anticipate head movements”. In: Journal of Neuroscience 26.34 (Aug. 2006), pp. 8838–8846.

[23] Avner Wallach et al. “Predictive whisker kinematics reveal context- dependent sensorimotor strategies”. In: (2019).

[24] Ben Mitchinson et al. “Feedback control in active sensing: Rat exploratory whisking is modulated by environmental contact”. In: Proceedings of the Royal Society B: Biological Sciences 274.1613 (Apr. 2007), pp. 1035–

1041.

[25] R. Bermejo, W. Friedman, and H. P. Zeigler. “Topography of whisk- ing II: Interaction of whisker and pad”. In: Somatosensory and Motor Research 22.3 (2005), pp. 213–220.

[26] Robert N.S. Sachdev, Takashi Sato, and Ford F. Ebner. “Divergent move- ment of adjacent whiskers”. In: Journal of Neurophysiology 87.3 (Mar.

2002), pp. 1440–1448.

Appendix A

State of the Art

A.1 Introduction to the background

Figure A.1: Principles for mysta- cial pad architecture. Reprinted from Brecht, Preilowski, and Merzenich [1]

with permission

The following is the “background chapter”, also called “State-of-the- Art chapter” or “literature study” be- longing to the master thesis “Analy- sis of mouse whisker movements and corresponding neuronal activity”. In this work, mouse whiskers will be tracked and analysed on their syn- chronicity.

A.2 Whisker anatomy

Location

Whiskers are stiff hairs which func- tion as a tactile organ found in var- ious mammals, like cats, seals, and rodents. The whiskers, also called vibrissae, can be subdivided based on their location into, for example, rhinal vibrissae (on the nose), labial vibrissae (on the lip), orbital vib- rissae (around the eye) and mysta- cial (on the moustache) vibrissae.

31

32 APPENDIX A. STATE OF THE ART

Brecht, Preilowski, and Merzenich [1] described some general rules for the distribution of the whiskers, illustrated in Figure 1. The anatomical terms of location are rostrocaudal (from head to tail) and dorsoventral (from back to belly). Regardless of species, most whiskers are located maximally rostral on the mystacial pad, in rows parallel to the rostrocaudal axis, with rows as far away from each other as possible, and arcs (columns) as close as possible to each other. In terms of orientation, the whiskers are perpendicular to the rostrocaudal axis and diverge within the dorsoventral plane of the row. Strik- ingly, the length of the whiskers increases exponentially in the rostrocaudal rows. The work in this thesis will be on whiskers of rats and mice.

Musculature

Dörfl [2] described the muscular anatomy of these mystacial vibrissae in mice.

He delineated five rows of follicles and named them A-E in dorsal-ventral di- rection and 1-7 along the row in caudal-rostral direction, with four follicles astride, named straddlers α-δ. Thus, this accounts for in total 31 whiskers growing from the individual follicles. The layout is schematically shown in Figure 2. Two types of muscles, namely intrinsic muscles and extrinsic mus- cles move the whiskers. The intrinsic muscles are wrapped like a sling around two consecutive hair follicles in a row. When contracted, they move the follicle backward. This, in turn, causes the connected whisker to protract (move for- ward). See Figure 3. There are four extrinsic muscles: m.levator labii superi- oris, m.maxillolabialis, m.transversus nasi, and m. nasalis. They are attached to the skull as well as the mystacial pad and are responsible for translation of the mystacial region medially backward, laterally backward, medially, and for- ward, respectively. See Figure 4. In resting position, the follicles are directed forwards, and the vibrissae are slightly retracted, due to the elastic properties of the connective tissue.

Haidarliu et al. [3] compared the muscular anatomy of the mystical vib- rissae in rats. Following a similar nomenclature, but with additional follicles D7 and E7, 33 follicles are identified. The organization is shown in Figure 5. The extrinsic rat muscles m.nasolabialis, m.maxillolabialis, m.nasolabialis superficialis, and pars orbicularis oris of the m.buccinatorius were identified.

All but the last are corresponding to the mouse extrinsic muscles, even though

they have separate names. In addition to these extrinsic muscles, the 5 distinc-

tive parts of m.nasolabialis profundus were described, corresponding to the

m.nasalis in mice. When these parts contract, the vibrissae will retract. The

authors also made a distinction between the whiskers in nasal (row A-B) and

APPENDIX A. STATE OF THE ART 33

(a) Schematic drawing of the intrinsic mus- culature of the left mystacial region of a mouse. I, infraorbital; L, labial; M, mysta- cial; R, rhinal; S, supraorbital. Asterisks indicate rare muscle bundles. Reprinted from Dörfl [2].

(b) Schematic drawing of two neighbour- ing mystacial follicles in the same row. R, rostral; C, caudal. F, follicular (intrin- sic) muscle; L, longitudinal muscular band formed by fibres of m. levator labii sup.

and m. maxillolabialis; B, fibrous band; P, plate; N, follicular nerve accompanied by an artery. Reprinted from Dörfl [2].

(c) Schematic drawing of the extrinsic musculature of the left whisker pad of a mouse. I, infraorbital nerve; L, m. leva- tor labii superioris; M, m. maxillolabialis;

N, m. nasalis; 0, orbit; T, m. transversus nasi; S, septum intermusculare. Reprinted from Dörfl [2].

(d) Spatial organization of the mystacial vibrissae of the rat. α-δ, the four caudal- most vibrissa follicles (straddlers); A–E, the five vibrissal rows; NS, nostril; NV, nasal vibrissae; R, rostral; V, ventral.

Reprinted from Haidarliu et al. [3] with permission from Wiley.

Figure A.2: Whisker anatomy

maxillary (row C-E) regions of the mystacial pad.