15th European Molecular Imaging Meeting
supported by:

To search for a specific ID please enter the hash sign followed by the ID number (e.g. #123).
*.ics

New Methods & Methodology in Neuroimaging I

Session chair: Cornelius Faber (Muenster, Germany); Charlie Demene (Paris, France)
 
Shortcut: PW06
Date: Wednesday, 26 August, 2020, 5:30 p.m. - 7:00 p.m.
Session type: Poster

Contents

Abstract/Video opens by clicking at the talk title.

245

Pharmacologically-induced dynamic changes in connectivity assessed by functional ultrasound and automatic scoring in awake mice

Claire Rabut1, Jeremy Ferrier3, Adrien Bertolo3, Bruno Osmanski3, Xavier Mousset3, Sophie Pezet1, Thomas Deffieux1, Zsolt Lenkei2, Mickael Tanter1

1 INSERM U1273, ESPCI Paris, CNRS UMR 7587, PSL Research University, Physics for Medicine, Paris, France
2 UMR_S1266 INSERM, University Paris Descartes, Laboratory of Dynamic of Neuronal Structure in Health and Disease, Institute of Psychiatry and Neuroscience of Paris, Paris, France
3 Iconeus, Paris, France

Introduction

The cholinergic system is altered in most neurodegenerative disorders such as Alzheimer's or Parkinson’s diseases. It was proposed that its acute modulation might be reflected as altered resting-state functional connectivity (FC)[1]. However, most brain imaging modalities require sedation which can alter metabolism and neurovascular coupling and could bias FC outcome. In this proof-of-concept study, we used functional ultrasound[2][3] to investigate the modulatory effects of Scopolamine (ScoP) - a muscarinic receptor antagonist - on functional brain connectivity in awake and behaving mice.

Methods

Male C57BL/6 mice were anesthetized using 1.5% isoflurane + medetomidine infusion (0.1 mg/kg/hr). A metal frame was first fixed on mice skull to allow fixation of the ultrasonic probe. Mice were then trained to stay 4-5hrs in a custom mobile cage with the probe and imaging was performed on day 10.
ScoP was administered at 0.2, 0.5 or 3mg/kg  s.c after 10min of baseline acquisition. Reversal was studied by the injection of milameline, a muscarinic acetylcholine receptor agonist and a control group by injection of the peripherally-restricted methyl-scopolamine. After acquisition, stationary epochs were concatenated to assess FC matrix at rest.
We trained an automatic classifier to score treated and non-treated mice and evaluated it on the different groups using independent mice.

Results/Discussion

The classifier model pointed out an important increase of the hippocampo-cortical connectivity induced by ScoP injection on independant datasets. The classifier estimated a pathological score bellow 0.1 for all the baseline states, and led to a significant increase of this score after scopolamine injections at doses 0.5 mg/kg and 3 mg/kg.
After reversal with medetomidine, a significant decrease of the score toward the baseline was observed.
Methyl-scopolamine, which does not cross the blood brain barrier, on the other hand didn’t induce any significative increase of the pathological score according to the model. This control measures therefore demonstrates the neuronal and non-behavioural character of the alterations measured under scopolamine.

Conclusions

In conclusion, we introduced pharmaco-fUS as a simple modality for the study of drugs effects on the brain without anesthesia bias. We showed in this proof of concept study that the method is highly sensitive to alterations of the functional connectivity and can be used to automatically score the state of an animal and as such be used as a simple functional readout to investigate the effect and dynamics of drugs on the central nervous system.

AcknowledgmentMT, ZL and TD are cofounders and shareholders of the Iconeus company. Parts of the funding was provided by a funding from the Human Brain Project, Projet FUSIMICE ANR-15-HBPR-0004
References
[1] Shah et al, 2015, 'Acute modulation of the cholinergic system in the mouse brain detected by pharmacological resting-state functional MRI.', Neuroimage. Apr 1;109:151-9
[2] Deffieux et al, 2018, "Functional ultrasound neuroimaging: a review of the preclinical and clinical state of the art.", Curr Opin Neurobiol. Jun;50:128-135
[3] TIran et al,  2017, ‘Transcranial Functional Ultrasound Imaging in Freely Moving Awake Mice and Anesthetized Young Rats without Contrast Agent’, Volume 43, Issue 8, Pages 1679–1689
Keywords: functional ultrasound, mouse, pharmacology, resting state, functional connectivity
246

Brain wide BOLD oscillations in long-lasting fMRI measurements in rat occur upon both sensory stimulation and resting state

Henriette Lambers1, Lydia Wachsmuth1, Ping Zheng1, Cornelius Faber1

1 University Hospital Muenster, Translational Research Imaging Center (TRIC), Muenster, Germany

Introduction

Arterial vasodilation caused by cerebral vasomotion has been detected in different species1 and can cause oscillations of the blood flow2. These oscillations constitute a physiological noise source for T2* weighted fMRI measurements and may affect the network analysis of resting state data. However, they do not affect BOLD maps of stimulation measurements3. To evaluate the influence of the oscillations, we performed BOLD measurements with high temporal resolution of 100 ms upon sensory stimulation and in resting state.

Methods

We acquired BOLD fMRI measurements on 7 medetomidine sedated and ventilated Fischer rats using a 9.4 T small animal MRI and a 1 cm surface coil. Single-shot GE-EPI measurements (TR: 100 ms, resolution: 325 µm x 350 µm x 1.2 mm, 1 slice, bregma: 1 mm, duration: 5 minutes) were performed at different times after the start of medetomidine injection upon electrical stimulation of one forepaw (9 Hz, 1 ms-pulses, 1.5 mA) using 2 different paradigms (ON/OFF: 4/26 s, 5/25 s). Additionally, 10 minutes resting state measurements were performed before and after the stimulation experiments in 4 animals.
Using MATLAB, BOLD time courses of 4 ROIs (Fig. 1A) were extracted and a Fast Fourier Transformation was performed. Measurements with oscillations show a clear peak in the spectrum (Fig. 1B).

Results/Discussion

None of 7 stimulation datasets taken in the first 3 hours after the begin of medetomidine sedation showed oscillations. 16 of 20 stimulation measurements performed after the first 3 hours of medetomidine sedation showed oscillations. The average onset time was calculated as 3.4 ± 0.5 hours for stimulation measurements. The resting state datasets also showed almost no oscillations in the first 3 hours (1 of 7), while 7 of 9 datasets taken after more than 3 hours showed oscillations. Oscillations occurred in all brain regions. However, analysis was only performed in those regions indicated in Fig. 1A, as other regions such as S1up and CPu, were too noisy, due to their distance from the coil. The mean oscillation frequency was 0.18 ± 0.03 Hz. There was no difference between the regions examined. Stimulation data showed slightly higher frequencies (0.19 ± 0.02 Hz) than resting state data (0.17 ± 0.04 Hz). However, these differences were not significant according to a U-test (Fig. 2B).

Conclusions

BOLD oscillations appear after 3-4 hours of anaesthesia and have to be considered for both sensory stimulation and  resting state experiments lasting longer than 3 hours. Oscillations occur brain wide and may thus influence network analysis of fMRI data, while detection of local activation in BOLD maps is not compromised. Further analysis whether correcting for oscillatory signal fluctuation alters results of network analyses is ongoing.

References
[1] Aalkjær, C, Boedtkjer, D, Matchkov, V 2011, ‘Vasomotion – what is currently thought?’, Acta Physiol, 202, 253-269
[2] Kim, JH, Ress, D 2016, ‘Arterial impulse model for the BOLD response to brief neural activation’, NeuroImage ,124, 394-408
[3] Lambers, H, Wachsmuth, L, Albers, F ,et al. 2019, ‘fMRI signal oscillations in rats can be induced by electric paw stimulation but do not deteriorate the observed BOLD peak’, ISMRM , Abstract No. 3764
Fig. 1
 (A) The 1 mm surface coil was placed over the S1Fl region that was activated by the electrical paw stimulation. 4 ROIs were investigated: S1Fl ipsi- (blue) and contralateral (orange), M1 ipsilateral (purple), the last ROI includes the superior sagittal sinus (red). (B) Exemplary time courses and spectra of one measurement without (B1) and one with oscillation (B2). The peak near 0.9 Hz corresponds to the respiration.
Fig. 2

Oscillations of resting state and stimulation datasets have been investigated in 4 ROIs. (A) Percentage number of resting state (left) and stimulation (right) measurements showing oscillations. n is the number of datasets split into early (first 3 hours) and late (after 3 hours). In late measurements, oscillations occurred in all regions. (B) Oscillation frequency and standard deviation. The frequency does not differ between the different regions. Frequencies of resting state and stimulation datasets show small non-significant differences. Indicated p-values were determined using a U-test.

Keywords: BOLD oscillations, fMRI, sensory stimulation, resting state, rodent
247

Phase-Contrast X-Ray Tomography of Marmoset Cochlea

Jannis Justus Schaeper1, Marius Reichardt1, Marina Eckermann1, Jasper Frohn1, Christoph Kampshoff2, Tobias Moser3, Tim Salditt1

1 University of Göttingen, Institute for X-Ray Physics, Göttingen, Germany
2 Max-Planck-Institute for Experimental Medicine, Göttingen, Germany
3 University Medical Center, Department of Otolaryngology, Göttingen, Germany

Introduction

The cochlea is the receptor organ in the inner ear that transduces sound into neuronal activity. Both fundamental aspects of signal transduction and neuro-physiology as well as biomedical research (implant technology, hearing loss and disorders) requires three-dimensional (3D) imaging techniques capable to quantify the micro-anatomy.
We present 3D imaging of excised small-animal cochleae by phase-contrast x-ray tomography using highly brilliant synchrotron radiation, and show how this technique can complement classical histology and light sheet microscopy in a correlative imaging approach.

Methods

Classical CT offers weak contrast for soft tissue, since contrast formation is based on absorption only. For 3D imaging of weakly absorbing samples the phase-shift induced by the sample is used. The resulting intensity variations in the detector plane is decoded by phase retrieval based on the contrast transfer function (CTF). Before tomographic reconstruction, a ring removal algorithm is applied. We used a multiscale-setup at our endstation GINIX at DESY [1], allowing imaging with different field-of-views (FOV).
The reconstructed tomograms are segmented using both algorithms depending on machine learning and blob detection. Additionally, lightsheet microscopy has been performed on decalcified cochlea. X-ray tomograms with different FOVs and lightsheet microscopy data are registered.

Results/Discussion

In our presentation, we show how high contrast for soft tissue [2] can be achieved, in particular using our endstation GINIX at DESY [1], and how the phase-shift induced by the sample can be decoded from intensity variations in the detector plane based on phase retrieval algorithms.
In the parallel beam and a short propagation distance, we reach an effective pixelsize of 650 nm and a FOV of 1.3 mm. Using a widened beam by KB focusing and a long propagation distance, we can cover the entire cochlea with an effective pixelsize of 3.25 µm and a FOV of 6.7 mm.
In view of age-related hearing loss we particularly aim at quantitatively evaluating the number of spiral ganglion neurones (SGNs) and hair cells in different age groups of marmosets.
Importantly, the data offers high contrast and little noise, allowing automated segmentation.
The CT-images are compared to lightsheet microscopy data, making it possible to also infer information about structural changes induced by the clearing process.

Conclusions

Without extensive sample preparation involving removal of surrounding bone or decalcification, we show that the shape, volumes and densities of individual neurons can be assessed. In the analysis, age-dependent changes in the quantity of SGNs can be determined. Further analysis of the 3D data is in progress and will be published [3]. Next, we will also explore whether the approach can be translated to in-house mµ-CT  using a liquid-metal jet.

References
[1] T. Salditt et al, 2015, Compund focusing mirror and X-ray waveguide optics for coherent imaging and nano-diffraction, J. Synchrotron Radiat. 22, 867-878
[2] M. Töpperwien et al, 2018, Propagation-based phase-contrast x-ray tomography of cochlea using a compact synchrotron source, Sci. Rep. 8, 4922
[3] J. Schaeper, M. Reichardt, M. Eckermann, J. Frohn, C. Kampshoff, T. Moser et T. Salditt, in preparation
Keywords: phase-contrast, x-ray tomography, cochlea, segmentation
248

Simultaneous 3 and 2 photon microscopy monitoring of exosomal uptake by neuron cells in mouse brain.

Maria Kefalogianni1, 2, Tsakani Edisona3, 4, Nenedaki Electra3, 4, Xydias Dionisis1, 2, Lemonis Andreas2, Gkirtzimanaki Katerina4, Sotiris Psilodimitrakopoulos2, George Garinis3, 4, Manolis Stratakis2, 5

1 University of Crete, Department of Physics, Heraklion, Greece
2 Foundation for Research and Technology-Hellas, Institute of Electronic Structure and Laser, Heraklion, Greece
3 University of Crete, Department of Biology, Heraklion, Greece
4 Foundation for Research and Technology-Hellas, Institute of Molecular Biology Biotechnology, Heraklion, Greece
5 University of Crete, Department of Materials Science and Technology, Heraklion, Greece

Introduction

Neurodegeneration is characterized by inflammation in the brain and neuron cell death. Exosomes are potent mediators of intracellular communication in the brain and their trafficking is of major clinical importance(1). To study this delicate system we used deep imaging on thick acute brain slices. Simultaneous 3 and 2 photon  excitation fluorescence (3 and 2PEF) microscopy(2) was utilized for deep, non-invasive imaging at depths of ~200um. Thus, we were able to monitor how brain resident immune cells send small death signals to neuron cells in freshly prepared tissue specimens.

Methods

Multiphoton imaging was carried out using a custom-built raster-scanning microscope (Fig. 1). Light from a diode-pumped Yb:KGW fs oscillator (1030nm, 70fs, 80MHz, 1W) was inserted into an inverted microscope after expanded by a telescope and passed through a pair of galvanometric mirrors. The generated 3 and 2PEF signals were collected  in the backward (epi-) detection geometry(3)(4)(5), with the same objective lens (Plan-Apochromat ×20/0,8N.A) used for excitation. Then they were  filtered by a short-pass filter (680/SP, Semrock), and  split by 2 dichroic mirrors and 3 bandpass filters and were guided with appropriate optics into 3 PMTs. The filters used were 527/20, 595/31, 458/64 for fluorescent dyes PKH67 (2PEF), Alexa Fluor 555 (2PEF) and DAPI (3PEF).

Results/Discussion

We have developed deep imaging multiphoton microscopy to address whether the exosomes are uptaken by a specific neuron cell type as deep as ~200μm, exceeding the penetration brain imaging limit of a few tens of μm using common confocal microscopy. Specifically, in order to understand how aged brain immune system drives neurodegeneration, we used acute brain slices of a mouse model of accelerated microglia damage accumulation and analysed how their exosomal secretion affects neuron cell fate with simultaneous 2 and 3PEF microscopy. Then, the resulted image was acquired by computing the maximum (Fig. 2) of all of intensity images for the z-scan stack at the depth of ~200μm. The colours that are depicted to the following image correspond to the different fluorescent dyes mentioned below. Neuron cells are depicted with red (Alexa Fluor 555), exosomes with green (PKH67), and nuclei of cells with blue (DAPI), while the uptake of exosomes by neuron cells with orange colour.

Conclusions

This study demonstrates the use of simultaneous 3 and 2PEF imaging technique as a reference for deep, non-invasive imaging at large depths of brain tissue to monitor the communication between different brain resident cell types. 3PEF was used to image DAPI (cell nuclei) and it was complemented by the 2PEF microscopy of PKH67 (exosomes), and Alexa Fluor 555 (neuron cells).

Acknowledgment

We would like to acknowledge Aggeliki Belli for the sample preparation and Prof. Kiki Sidiropoulou for fruitful discussions. This work was financially supported by project EPIGRAPH:GRAPHene biomolecular and electrophysiological sensors integrated in an “all-in-one device’’ for the prediction and control of epileptic seizures (towards a general device for most brain disorders). Sotiris Psilodimitrakopoulos acknowledges financial support from IQONIC-EU, H2020, under Grant Agreement N 820677. Gkirtzimanaki Katerina acknowledges financial support from ERC GA646663 DeFiNER.

References
[1] Saeedi, S., Israel, S., Nagy, C. et al. , 2019, Transl Psychiatry 9, 122, pp. 2158-3188.
[2] Guesmi, K., Abdeladim, L., Tozer, S. et al., 2018, Light Sci Appl 7, 12, pp. 2047-7538.
[3] Psilodimitrakopoulos, S., Mouchliadis, L., Paradisanos, I. et al. , 2018, Light Sci Appl 7, 18005, pp. 2047-7538
[4] Psilodimitrakopoulos, S., Mouchliadis, L., Paradisanos, I. et al. , 2019, Sci Rep 9, 14285, pp. 2045-2322.
[5] Maragkakis G M, Psilodimitrakopoulos S, Mouchliadis L, Paradisanos I, Lemonis A et al., 2019, Opto-Electron Adv 2, 190026 , pp. 190026.
Schematic Representation of Setup

Figure 1: Schematic representation of the experimental setup. Abbreviations, as met by the laser fundamental pulse: L: lens, OL: Objective Lens, GM: galvanometric mirrors, M: mirror, DM: dichroic mirror, BF: Bandpass filter, SF: Short pass filter, PMT: photomultiplier tube.

Maximum intensity of brain slice at depth of 200microns.
Figure 2 : Maximum intensity of brain slice at depth of 200microns. With red colour are the neurons that have been labelled by the Alexa Fluor 555 (2PEF) antibody, while with green are the exosomes that have been pre-labeled with PKH67 fluorescent dye (2PEF). Finally, with the blue colour are the nuclei of cells that have been labeled by the fluorescent dye DAPI (3PEF). Scalebar shows 20μm.
Keywords: simultaneous 3PEF and 2PEF microscopy, brain slice, neurodegeneration, exosomes, neuron cells
249

Characterization of haemodynamic responses to peripheral sensory stimulation in mice with functional optoacoustic and ultrasound imaging

Justine Robin1, 2, Richard Rau3, Berkan Lafci1, 2, Aileen Schroeter1, Michael Reiss1, 2, Xosé L. Dean Ben1, 2, Orçun Göksel3, Daniel Razansky1, 2

1 ETH Zurich, Institute for Biomedical Engineering and Department of Information Technology and Electrical Engineering, Zurich, Switzerland
2 University of Zurich, Faculty of Medicine and Institute of Pharmacology and Toxicology, Zurich, Switzerland
3 ETH Zurich, Computer-Assisted Applications in Medicine (CAiM) Group, Zurich, Switzerland

Introduction

Functional imaging of the whole brain is key to understand the complex mechanisms underlying brain function. If fMRI remains a gold standard, functional ultrasound (fUS) is now a well-established tool to measure cerebral blood volume (CBV) response[1]. Functional optoacoustic neuro-tomography (FONT) has recently enabled measuring other hemodynamic responses (oxy–(HbO), deoxy-(HbR) and total hemoglobin (HbT)) in mice[2]. Here we compare fUS and FONT responses to electrical paw stimulation in mice, and comment on their complementary value for understanding the neurovascular coupling mechanisms

Methods

Anesthetized mice (n=4, protocol adapted from [3]) were imaged with fUS and FONT during electrical stimulation (5Hz,0.5mA,15s,random interval) provided in an alternate fashion via needle electrodes placed in both hindpaws. A linear probe (L22-14v, 16MHz) and a 128-ch. ultrasound Vantage system (Verasonics) were used for fUS. Cross-sectional power Doppler images (300 frames) were obtained with plane wave compounding (20 angles, 1kHz), GPU-based beamforming and SVD filtering[4].A 512-channels spherical probe (150°,7 MHz, Imasonic) and acquisition system (Falkenstein Mik.) were used for FONT. Multispectral data were acquired with a wavelength-tunable laser (530, 540, 560, 575, 585nm, 25Hz, Innolas) and unmixed for HbO, HbR and HbT to reconstruct 10x10x4mm3 volumes (model-based algorithm [5]).

Results/Discussion

As shown in Fig. 1, fUS imaging provided a high quality 2D vascular map up to 10 mm deep in the tissue. Upon electrical hindpaw stimulation, a response was observed in the contralateral somatosensory hindlimb region (S1HL), while the ipsilateral signal in S1HL remained at baseline level. The change in CBV was around 2%, with a peak occurring 10 s after the end of the stimulus presentation. These results closely match those previously reported in fMRI in a similar study[3].
As shown in Fig. 2, FONT provided high quality 3D vascular mapping of the mouse brain, but with a reduced penetration depth of about 4 mm due to the strong attenuation of visible light wavelengths in tissue. Upon stimulation, a response was observed both in HbT and HbO (respectively Fig. 2-a, and Fig. 2-b). The evoked HbO signal increase was between 15 and 20%, with a peak occurring up to 30 s after the end of the stimulus presentation. HbT signal increase was between 5 and 10%, with a peak at ~20s after stimulus end.

Conclusions

Our stimulation protocol reliably induced a specific activation in the contralateral S1HL area readily identified with both fUS and FONT. Interestingly, though the activated regions match, the CBV and HbT responses differ in amplitude and temporal length. This may give finer insights on the mechanisms of the vascular response to neural activation.Further studies will aim at characterizing dependence of the induced responses on stimulus parameters

AcknowledgmentRichard Rau and Justine Robin contributed equally to this work
References
[1] T. Deffieux, C. Demene, M. Pernot, and M. Tanter, “Functional ultrasound neuroimaging: a review of the preclinical and clinical state of the art,” Current Opinion in Neurobiology, vol. 50. Elsevier Ltd, pp. 128–135, 01-Jun-2018.
[2] S. Gottschalk, T. Felix Fehm, X. Luís Deán-Ben, and D. Razansky, “Noninvasive real-time visualization of multiple cerebral hemodynamic parameters in whole mouse brains using five-dimensional optoacoustic tomography,” J. Cereb. Blood Flow Metab., vol. 35, pp. 531–535, Mar. 2015.
[3] H. J. Shim et al., “Mouse fMRI under ketamine and xylazine anesthesia: Robust contralateral somatosensory cortex activation in response to forepaw stimulation,” Neuroimage, 2018.
[4] R. Rau et al., “3D functional ultrasound imaging of pigeons,” Neuroimage, vol. 183, pp. 469–477, Dec. 2018.
[5] L. Ding, X. L. Dean-Ben, and D. Razansky, “Efficient 3-D model-based reconstruction scheme for arbitrary optoacoustic acquisition geometries,” IEEE Trans. Med. Imaging, vol. 36, no. 9, pp. 1858–1867, Sep. 2017.
Figure 1
Top - PW Doppler image of the mouse brain. Overlaying are the mean CBV increase upon (a) Right hindpaw stimulation and (b) Left hindpaw stimulation. Bottom – CBV responses in the right (blue) and left (green) S1HL ROI upon (c) Right hindpaw stimulation and (d) Left hindpaw stimulation
Figure 2
Top – Volumetric optoacoustic image of the unmixed total haemoglobin (HbT). (a) Overlaying are the mean HbT signal increase upon right hindpaw stimulation (blue) and left hindpaw stimulation (green). Bottom – (c) metabolic responses of the oxy- (red), total – (black) and deoxy- (blue) haemoglobin in the right S1HL ROI upon right hindpaw stimulation. (d) metabolic responses of the oxy- (red), total – (black) and deoxy- (blue) haemoglobin in the left S1HL ROI upon left hindpaw stimulation
Keywords: Optoacoustic tomography, neuroimaging, functional ultrasound
250

DIY: 2nd order Human Brain Atlases

Claudiu Ivan1, Andrea Mendez-Torrijos1, Marina Sergeeva1, Andreas Hess1

1 Friedrich-Alexander-Universität Erlangen-Nürnberg, Institute of Experimental and Clinical Pharmacology and Toxicology, AG HESS, Erlangen, Germany

Introduction

Given the extensive offer of human brain atlas, which´s parcellations depend strongly on the author’s preferences, no available atlas contains and will every single brain parcellation. We found a lack of flexibility to adapt existing brain atlases to precise scientific goals. Moreover, there is increased awareness of the need for disorder specific atlases for tailored analyses. Relevant brain structures are often ignored due to their unavailability in the atlas of choice, existing however, in other atlases. We will exemplify through the creation of a specific atlas for Anorexia Nervosa (AN).

Methods

After the target brain regions are chosen, the different atlases have to be adapted and combined to become a study-specific optimal “multi-atlas”. Using the Harvard-Oxford atlas [1] as main one we decided to add AN relevant subcortical and cerebellar structures from other 6 atlases. For this we propose a workflow (figure 1) as a decision-tree according to each atlas unique properties. Here we suggest a method for creation of probabilistic atlases from non-probabilistic brain structures. Further, we describe different registration ways to the common template space of choice [2, 3]. During this procedure, every node can belong to different brain regions, however, the joint probabilities should never exceed 100%. To overcome this we define different normalization approaches.

Results/Discussion

The workflow and methodologies described result in a more comprehensive atlas that can be constantly re-tailored to specific goals. Nevertheless, certain factors should never be ignored when considering atlas implementation such as probabilistic properties, initial segmentation or size of the brain structures. Every atlas does come with limitations, such as the different parcellations were done on different subjects/templates, which can result in inconsistent probabilistic distributions for each brain region. Given that till the date there is not an extensive enough atlas this is approach offers an acceptable compromise.

Conclusions

Currently, there are no generalizing methodological solutions for goal-directed multi-atlas generation. We present an effective workflow that allows scientists to generate tailored atlases by adapting and combining different atlas resources. Moreover, the workflow allows to depict disorder specific atlases which would contribute to improve, both functional and anatomical analysis.

Acknowledgment

We thank our work group AG HESS at the FAU Erlangen-Nürnberg and specially Silke Kreitz.

References
[1] In Harvard-Oxford Cortical and Subcortical Structural Atlases. http://www.cma.mgh.harvard.edu/.
[2] Paul A. Yushkevich, Joseph Piven, Heather Cody Hazlett, Rachel Gimpel Smith, Sean Ho, James C. Gee, and Guido Gerig. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage 2006 Jul 1;31(3):1116-28.
[3] Avants, B. B., et al. (2011). "A reproducible evaluation of ANTs similarity metric performance in brain image registration." Neuroimage 54(3): 2033-2044.
Multi-atlas creation workflow
The figure ilustrates the workflow and methodologies described for the creation of the goal-directed multi-atlas.
Keywords: Brain atlas, atlas creation, Multi-atlas
252

A novel user-friendly smartphone ophthalmoscopy method to track retinal vasculature changes in a mouse model of ischemic stroke

Marin Radmilović2, Anja Barić1, Helena Justić1, Jelena Kežić1, Siniša Škokić1, Antun L. Brkić3, Marina Dobrivojević Radmilović1

1 University of Zagreb School of Medicine, Croatian Institute for Brain Research, ZAGREB, Croatia
2 Sestre milosrdnice University Hospital Center, Department of Ophthalmology, ZAGREB, Croatia
3 Institute of Physics, ZAGREB, Croatia

Introduction

Transient intraluminal occlusion of the middle cerebral artery (MCAO) in rodents leads to simultaneous induction of cerebral and retinal ischemia. The aim of the current study was to establish a simple and affordable novel fluorescein angiography and fundus photography method, which would allow us longitudal in vivo monitoring of mouse retinal vasculature changes after stroke induction.

Methods

Brain and retinal ischemia were induced in C57Bl6J mice by a 30-minute MCAO. The animals were scored for neurological deficit and imaged by MRI. Our ophthalmoscope captures a digital image of the fundus in the smartphone camera attached to a stereo microscope and aligned with the ocular lens, allowing the visualization of the entire lens diameter. The anesthetized mouse is secured below the microscope in a mobile holder. Both eyes are treated with Mydriacyl for pupils dilation and covered with Recugel to prevent drying of the cornea. The eye is positioned in the focal plane of the microscope and a hand held 90D lens is maneuvered over it until the optimal image is captured. For fluorescein angiography the same setup is used with the addition of an excitation filter and a barrier filter.

Results/Discussion

Retinal fundus images are usually obtained using expensive and large tabletop clinical systems. Here we report a novel low cost and user-friendly smartphone fundus photography and fluorescein angiography method with the accompanying image analysis pipeline to monitor the changes in the retinal vasculature. To analyze the retinal vascular area, retinal venous and arterial width, arterial tortuosity, arterial branching and arteriovenous shunting the high resolution images were pre-processed using the Perona-Malik anisotropic diffusion equation to maximize the signal to noise ratio. Using a pre-trained neural network, we determined the position of the optic disc and its perimeter from which the starting positions of each arterial and venous branch were defined. Our algorithm allows tracing the vascular trajectory reconstructing the vascular skeleton, which is further mathematically processed for width, length, curvature and angle of intersection.

Conclusions

We developed a single setup for both fundus photography and fluorescein angiography with high quality resulting images and the accompanying image analysis algorithm which is able to detect and quantify the changes in the retinal vasculature after MCAO. Our smartphone ophthalmoscope allows us to monitor the retinal vascular status over a longer period giving us an insight in the pathophysiological changes that occur following ischemic stroke.

AcknowledgmentThe study is supported by the Croatian Science Foundation project BRADISCHEMIA (UIP-2017-05-8082). The work of doctoral student Anja Barić has been fully supported by the “Young researchers' career development project – training of doctoral students” of the Croatian Science Foundation funded by the European Union from the European Social Fund. Multimodal imaging was done at Laboratory for Regenerative Neuroscience - GlowLab, University of Zagreb School of Medicine.
References
[1] Allen RS, Sayeed I, Cale HA, et al. Severity of middle cerebral artery occlusion determines retinal deficits in rats. Exp Neurol. 2014;254:206–215.
Kamalaveni V, Anitha Rajalakshmi R, Narayanankutty KA. Image Denoising Using Variations of Perona-Malik Model with Different Edge Stopping Functions, Procedia Computer Science. 2015;58:673-682.
Li H, Lin Z, Shen X, Brandt J and Hua G. A convolutional neural network cascade for face detection. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, 2015;5325-5334.
Keywords: smartphone ophthalmoscope, fluorescein angiography, stroke