1

My name is Praveen Pankajakshan, ...

Hello!

2

Y

... and this is my

3 Powered by: tagul.com

Biography Cloud

4

Academics

5

IIT ROORKEEINDIA

UNDERGRADUATEElectrical Engineering

6

Texas A&M University

College StationUSA

GRADUATE SCHOOL Electrical & Computer Engineering

7

INRIA Sophia Antipolis&

University of Nice-Sophia Antipolis

PHD SUMMA CUM LAUDEComputer Science

8

PhD Thesis

Jury ReportCORDI-s Fellowship

9

What I do now ...

Post Doctoral Fellow

10

PhD+Postdoc research cloud

Powered by: academia.edu

11

x

Where I live ...

Expertise: Computational Nanoscopy

12

For me research is like fine cuisine ...

13

ToolsCreativityObjectivity

14

My Experiences

are drawn from

15 Powered by: tagul.com

My diverse research cloud...

16

Coll

abor

ativ

eR

esea

rch

Wha

t I b

elie

ve in

...

17*P. Pankajakshan, et al. Deconvolution and Denoising for Confocal Microscopy. In F.Cazals and P. Kornprobst, editor, Modeling in Computational Biology and Biomedicine, ch.4, Springer, In Publication, 2012.

Upcoming book chapter*

2012+

12 peer reviewedconference and journal

articles+

many invited talks

18

Biography listed in

Who’s who in the World2010, 2011, 2012

PSF model for fluorescence MACROscopy imagingPraveen Pankajakshan1, Alain Dieterlen2, Gilbert Engler3, Zvi Kam4, Laure Blanc-Féraud5, Josiane Zerubia5 & Jean-Christophe Olivo-Marin1

1Quantitative Image Analysis Unit, Pasteur Institute, France,2Laboratoire MIPS-LAB.EL, Universite de Haute-Alsace, France.

3IBSV Unit, INRA, France,4Molecular Cell Biology, Weizmann Institute of Science, Israel.

5Ariana joint research group, INRIA/CNRS/UNS, France.

INTRODUCTION

Fluorescent MACROscope is useful for observing large samples (of the order of a few centimeters) and has the following advantages:

• large object fields, • large working distances, and • parallax-free imaging.

Imaging field aberrations using point source

CONCLUSIONS

1. MACROscope PSF varies as a function of the lateral position. 2. Vignetting was observed for small zooms (large FOV). 3. This initial PSF model will be enhanced with newer

acquisitions on different systems.

www-syscom.univ-mlr.fr/ANRDIAMOND/www.bioimageanalysis.org{praveen,jcolivo}@pasteur.fr

alain.dieterlen@uha.fr

OBJECTIVE

METHODOLOGY

Why A PSF MODEL IS IMPORTANT?

Unknown synthetic object

PSFObserved volume

Results

Experimentally obtained bead

images for different lateral

position in the field

X

X X

X X

To model the point-spread function (PSF) of a MACROscope operating with field aberrations due to optical vignetting.

ACKNOWLEDGEMENTSThe first author wish to thank ANR DIAMOND for funding the postdoctoral research fellowship. The authors also are grateful to Dr. Philippe Herbomel from the Institute Pasteur, France and Dr. Didier Hentsch from IGBMC, France for the images and the discussions.

Telecentric lens assembly

Pupil function for a MICROscope

Pupil function for a MACROscope

SIMULATE PUPIL OPTICAL VIGNETTING AND PSF

Cat’s eye effect

Simulated PSF

Measured bead image

{

{

Lens displacement x-direction

Lens displacement y-direction

Point-spread function (PSF)-> Image of ideal point source (4 m).Total magnification 6.25x.

XY

ZXY

Z

XY

Z

XY

Z

XY

Z

XY

Z

MACROscope lens assembly

Stokseth’s PSF model Excitation PSF Emission PSF

19

MiFoBio’10

Best Poster Award

from Nikon

YRLS’10

Public Poster Award

20

If you like my work and have a

X

21

Praveen Pankajakshan

15 rue de l’ésperance

75013 Paris

France

www.linkedin.com/in/praveenpankaj

+33 (0) 628358064

www.bioimageanalysis.org/~praveen

praveen.pankaj@gmail.com

pasteur.academia.edu/PraveenPankajakshan

praveenpankaj

Contact me ...

22

Research Highlight

Deconvolution as virtual inverse lens

23

Observ

ed

Deconvo

lved

Application: Fluorescence Microscopy

24

Widefield (WFM) Confocal (CLSM)

M. Minsky. Memoir on inventing the confocal scanning microscope. Scanning, 10:128–138, 1988.

25

Convallaria rhizome (Courtesy: INRA)

Volume onWFM

Volume onCLSM

Background Fluorescence!

Better but ...Low photon count!

26

Frequency Spectra for Convallaria sample

WFM CLSM

High frequencies lost!

... and yet, there is the Diffraction Barrier

27

Minimum Radial Resolution ~ 200nm

d = 0.61λ

NA

28

but how?

SNR

ResolutionSpeed

Acquisition

I minimize trade-offs in microscopy ...

... by bridging optics and image Processing

• Improvement in resolution.

• Noise reduction

? ...but requires point-spread function (PSF) h(x).

29

MicroscopeSample 3-D Image

Diffraction limit Aberrations Noise

Inverse approach

?

30

Complexity

Identify Point Sources

FluorescentBeads

Blind estimation

Separate PSF estimation

Point-Spread Function (PSF) h(x)

Accuracy

PSF Model

PSF Modeling

31

Fourier Transform

Defocus Pupil

Function

AmplitudePSF

Squared Magnitude

MicroscopePSF

P. A. Stokseth, “Properties of a defocused optical system,” J. Opt. Soc. Am. A, vol. 59, pp. 1314–1321, Oct. 1969.

Clear Pupil

Apodized Pupil

P. Pankajakshan et al. “Blind Deconvolution for Confocal Laser Scanning Microscopy,” PhD Thesis., Dec. 2009.

!22

!20

!18

!16

!14

32

Axial MIP PSF for a 40x/1.4 oil immersion lens.

!12

!10

!8

!6

WFM CLSM

z

x

z

x

P. Pankajakshan, et al. Deconvolution and Denoising for Confocal Microscopy. In F.Cazals and P. Kornprobst, editor, Modeling in Computational Biology and Biomedicine, ch.4, Springer, In Publication, 2012.

Determining PSF- Beads as ‘guide stars’

33

Imaging fluorescent beads

xy

z

Imag

e P

lan

eO

bjec

t P

lan

e

Hands on

experience!

34

Blind/Myopic Deconvolution

Update PSF

Update Object

Alternate PSF & Object estimation

Observation

New PSFNew

Object

Old Object

Old PSF

T. J. Holmes. Blind deconvolution of quantum-limited incoherent imagery: maximum-likelihood approach. J. Opt. Soc. Am. A, 9:1052–1061, July 1992.

B. Zhang et al. “A study of Gaussian approximation of fluorescence Microscopy PSF models,” SPIE conf., San Jose, Jan. 2006.

PSF constraints Object constraints

P. Pankajakshan et al. “Blind Deconvolution for Confocal Laser Scanning Microscopy,” PhD Thesis., Dec. 2009.

35

Observed

Restored

Convallaria cell from Parenchyma tissue

Focussed on the cytoplasmic threads of a cellP. Pankajakshan, et al. Deconvolution and Denoising for Confocal Microscopy. In F.Cazals and P. Kornprobst, editor, Modeling in Computational Biology and Biomedicine, ch.4, Springer, In Publication, 2012.

Blind Deconvolution of Fluorescent Shell

36

Observed

FWHM: 930nmThinBlinDe: 40 iter.

FWHM: 535.58nm

15 microns

Specification:Diameter

Thickness specified:

500-700nm

P. Pankajakshan et al. “Blind Deconvolution for Confocal Laser Scanning Microscopy,” PhD Thesis., Dec. 2009.

2x increasein resolution!

Non-Blind vs Blind

37

Commercial software(Using PSF model)

Proposed ThinBlinDe

38

Commercial Non-blind restoration

Convallaria cell from Parenchyma tissue

ThinBlinDe: Blind

restorationObserved

Courtesy: INRA

P. Pankajakshan et al. “Blind Deconvolution for Confocal Laser Scanning Microscopy,” PhD Thesis., Dec. 2009.

Thank you

39

40

Praveen Pankajakshan

15 rue de l’ésperance

75013 Paris

France

www.linkedin.com/in/praveenpankaj

+33 (0) 628358064

www.bioimageanalysis.org/~praveen

praveen.pankaj@gmail.com

pasteur.academia.edu/PraveenPankajakshan

praveenpankaj

Contact me ...