Deconvolution, Visualization and

Analysis Software for Widefield and Confocal

Microscopy

Media Cybernetics is committed to delivering the most advanced algorithms available on the

market today by the cutting edge research to develop the most advanced algorithms for

tomorrow.Media Cybernetics provides uncompromised

results with faster performance than ever before.

The best just keeps getting better and faster…

Deconvolution:

A crash course

Why Deconvolution ?

Each single plane contains blur from planes above and below

PSF: 3D representation you get when imaging a sub-resolution latex bead

Diffraction and aberration in the optical path lead to Blur / Haze

3-D diffraction pattern (= Point Spread Function PSF)

Blurred, faint

Sharp, bright

Blurred, faint

z

What is “Deconvolution” ?

“Deconvolution” is a computational procedure

that restores the quality of images.

Usage:

• Microscopy: biomedical research, ophthalmology, medical diagnostics, nuclear imaging (CT), food processing, forensics, skin care/beauty product development, etc.

• Astronomy: telescope imaging, satellite imaging, aerial imaging (cartography), etc.

• Industrial: Quality control for printed circuit boards and semiconductors, etc.

(Spirogyra)

Benefits of Deconvolution

1. Haze removal 1. Haze removal More contrast More contrast Better segmentation Better segmentation

2. Improved resolution 2. Improved resolution More detail revealed More detail revealed

3. Access to depth information3. Access to depth information

4. Produces quantitative results (non-destructive)4. Produces quantitative results (non-destructive)

Algorithms:

Deblurring or Deconvolution ?

Deconvolution vs. Deblurring

• Deconvolution– More closely approximates a

reversal of the blurring process

– Produces Quantitative results (reverses blurring accurately)

– Better representation of actual morphology

• Deblurring– Can make images look better

fast.– Results are strictly qualitative – Produces artifacts– Morphologically unreliable.

Qualitative vs. Quantitative

No-Neighbors Single slice haze removal.

Nearest-Neighbors Haze removal from a few optical sections.

Inverse filtering Inversion of the haze using a theoretical PSF.

Constrained iterative Iterative deconvolution with a fixed PSF.

Maximum likelihood Max likelihood estimate-based iterativedeconvolution with a fixed PSF.

Adaptive blind Max likelihood estimate-based iterativedeconvolution deconvolution deriving the PSF from

information found in the ‘haze’ itself.

No/Nearest Neighbors

• Uses center slice and two adjacent slices• Estimate amount of blur that adjacent slices contribute to center

slice• Subtract away the blur (remove photons)• Deblur one slice at a time• Fast, 2D, only qualitative (unsharp mask)

0

+z

-z

Inverse/Wiener Filter

• Restoration algorithm – looks at whole structure• Works well on large objects• Fast, 3D, qualitative• Regularization/Stability

ImageObject PSF

DeblurredImagePSF

1

Imaging:

Inverse Filter:

Constrained Iterative Deconvolution

• Requires PSF (theoretical or measured)• Iteratively improve object estimate • Impose constraints (non-negativity)• Requires highly optimized imaging system

ReblurredObjectEstimate

PSF Original Blurred

ImprovedEstimate

-

Update Error

Iterative Blind Deconvolution

• Estimate both PSF and Object!• Need microscope parameters (NA, RI, xyz, λ)• Theoretical PSF with spatial/spectral constraints• Object with non-negativity and noise constraints

ObjectEstimate Update

ObjectPSF

Estimate

UpdatePSF

OriginalBlurred

Why AQI and “Blind” Deconvolution ?

• Ease of Use– No need to image beads and measure PSF

• Much less expertise required by the user• Less time required on the microscope• Allows for processing of existing data• Runs on PC• Allows for deconvolution of large images• The reconstructed PSF can vary over the volume• Disadvantages

– More processing time

When to use what?

When Speed & Haze removal are important

No Neighbor

Nearest Neighbor

Inverse Filter

3D Deconvolution

When Quality & Quantitation are critical

XY Max Projections

XY Max Projections

XZ Max Projections

XZ Max Projections

Raw

InverseFilter

Nearest Neighbors

Blind Deconvolution

3D Blind Deconvolution

Drosophila larvae leg

3D Blind Deconvolution

Blind deconvolution – 3D Confocal

Blind deconvolution – 3D Confocal

2D Blind deconvolution – TIRF

Blind deconvolution (with Spherical aberration correction)

How do I know it’s real?

raw

deblurred

Structure from Structure from nothing?! nothing?!

slice 29

slice 37

raw deblurred

Optical Slices

2D Blind

2D Blind Deconvolution – true iterative blind deconvolution

• No image parameters required• Ideal for optically thin specimens

• Assumes everything is in one plane of focus

2D Blind Deconvolution2D deconvolution on a 3D volume

Why Deconvolve?

• Improved resolution of small features

• Better definition of structure in multiple dimensions

• More accurate localization of intensity for quantification and ratiometry

• Signal to noise (SNR) improvement

Why Media Cybernetics…

• Established technological focus in deconvolution – over 25

years of dedicated research

• Breadth of algorithms that are well known and accepted world-

wide – optimized for maximum robustness

• Media Cybernetics stands behind all of its products and

provides world class education and support.

• Committed focus to deconvolution, visualization and analysis

R&D ensures customer access to the latest technologies for

today and tomorrow.

Media Cybernetics:Offering complete post-processing solutions

• Image improvement, restoration, visualization and analysis that facilitates more accurate and compelling results.

• Media Cybernetics creates the tools to help you do the science.

• AQI Product Line– AutoDeblur (2D, Silver, Gold)– AutoVisualize (with improved 3D-Viewer)– AutoAnalyze (FRET, Ratio, Object Counting/Tracking,

Colocalization)– Tools (automatic, manual and channel-channel alignment)

• Image-Pro Product Line– Image-Pro Plus and Image-Pro Plug-In module bundles– Express– Analyzer