3D Image Deconvolution

Basic Microscopy Requirements

MagnificationHow large can I make something appear?

ResolutionHow small an object can I detect?

ContrastWhat is real and what is artifact?

Magnification

Resolution

DIC image, 100X magnificationA. North, JCB 2006

Contrast

Signal to BackgroundBackground is the systematic addition to the intensity of the signal.

Contrast

Signal to NoiseNoise is any random variation that adds to the uncertainty of the signal

Microtubules (Objects)

Microtubules (Images)

Appearance of a 100 nm (0.1 mm) Fluorescent Bead in a Light Microscope

Images from the beadBead

Point Spread Function (PSF)

y

x

z

x

y

x

Point Spread Function: measurement of a physical phenomenon

Convolution

ActualActual MeasuredMeasured

2 Adjacent Points2 Adjacent Points 4 Adjacent Points4 Adjacent Points

PSF in the Bead

PSF in the Sample

Confocal vs Widefield Microscopy

Pinhole

Pinhole

Confocal Widefield

Illumination

Detector

Object

Condenser

Objective

Confocal vs Widefield Microscopy

Confocal Widefield

Image Image

Deconvolution Microscopy

PSF Image

=

PSFImage

=-1

Object

Object

YFP-α-tubulin in T. gondii

Swedlow et al., PNAS 2002

Confocal Wide-field Deconvolution

Deconvolution

Accounts for “blur” introduced by the optical pathwayImproves contrast

Two types of deconvolution 1. Deblurring/Nearest Neighbor2. Image Restoration/Constrained Iterative

Comparison of Deconvolution Methods

XLK2 Cell: Ab: α-tubulin-Cy5

Deblurring

Uses 3 z planes: focal plane +/- 1 planeTreats a 3D image as a 2D image

ProsFastMay be visually pleasing

ConsInvolves a few assumptions:1. Out-of-focus light from other planes is negligible2. PSFs are equivalentSubtractive methodNot quantitative

Deblurring

Imagen = Objectn ⊗ PSFn + Objectn+1 ⊗ PSFn+1 + Objectn-1 ⊗ PSFn-1

Z

n0

n-1

n+1

Image Restoration

Uses all z planes collectedTreats a 3D image as a 3D image

ProsQuantitativeReassigns light to plane of originImproves signal-to-noise ratio

ConsRequires computational resourcesSensitive to optical imperfections

Image Restoration

• Guess at In• Convolve In with PSF• Compare to On

• Check for negative intensities• Make new guess• Continue until a stable guess is achieved

What is a guess?!?A “guess” is where the signal distribution in x, y, and z can

be calculated to satisfy the above equation

Imagen = Objectn ⊗ PSFn

Image Restoration

Z

Image Restoration Microscopy

Image Restoration Microscopy

Comparison of Deconvolution Methods

XLK2 Cell: Ab: a-tubulin-Cy5

Deconvolution

What are the system requirements for deconvolution?Deconvolution is not just about algorithms!

Blur is introduced by the optical pathwayYou must examine your opticsNot all PSFs are equal because…

Not all lenses are equalPrecise stage movement

Objectives

J. Murray, CSHL 2006.

Objectives

J. Murray, CSHL 2006.

Objectives

CLEAN THEM!

J. Murray, CSHL 2006.

X

Z

PSF Varies by Objective

Coma

P. Goodwin

Axial Skew

CAA B C

A B CA B C

Spherical Aberration

Chromatic Aberration

SDY = 0.041 mmSDZ = 0.037 mm

SDY

= 41 nmSD z = 37 nm

A pixel is 0.064 mmor 64 nm

OpticalResolution

~ 200 nm in XY~ 500 nm in Z

A pixel and the voxel are the smallest units that compose an image or a volume

Stage RepeatabilityDeviation from Median Position

61 Point Visits over 24 minutes

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

-0.5 -0.3 -0.1 0.1 0.3 0.5

Y Position (um)

Z Po

sitio

n (u

m)

OneVoxel 64 nm

Environmental Chamber

Maintain environment for cells: heat and humidity.Provides protection from dust, dirt and grime

Stage Drift• Mechanical Factors

– Overall stage stability– Z drift: motorized stage vs objective

• Environmental Factors– Room temperature– Humidity– Air conditioning and heating fluctuations

Why Bother?

If so many things can affect deconvolution, why bother?

QuantitativeIncreases signal-to-noise ratio: critical in live-cell

imagingHigh resolutionComputer processing speeds continue to improveWide-field imaging is gentle on live cellsMaximize your data collection while minimizing

phototoxicity

HIV-1 Vpr Nuclear Envelope Localization

Vodicka et al., Genes Dev. 1998

Vpr colocalizes to nuclear membrane pores

2.5 µm

Confocal and epifluorescent data indicated nuclear localizationBiochemically Vpr was membrane associated

HIV-1 Vpr Nuclear Envelope Localization

Vodicka et al., Genes Dev. 1998

1.0 µm

mab 414:recognizes subset of nucleoporins

HIV in Dendritic- T-Cell Synapses

McDonald et al., Science 2003

HIV viral particles recruited to synapse

DNA

ActinHIV

Protein trafficking

M. Falk, Scripps Research Institute

YFP-tubulinCx43-CFPlysosomes

One cell nucleus from one field of 40 in a 24 hour time lapse experiment

Histone-2B-GFP

L. Pelletier, Max Planck Institute

Tubulin-GFP

L. Pelletier, Max Planck Institute

Centrin-GFP

L. Pelletier, Max Planck Institute