3D light microscopy techniques
The image of a point is a 3D feature In-focus image Out-of-focus image
The image of a point is not a point Point Spread Function (PSF) 1D imaging 2D imaging 3D imaging
Resolution is now an arbitrary measure of how close two point images can come such that they are perceived as separate Lord Rayleigh s criterion: λ = 488 nm (NA = 1.4) δ = 212 nm; δ z = 780 nm (NA = 0.4) δ = 744 nm; δ z = 9.56 micron
Image formation in a light microscope Φ( x) = Ψ( y) PSF( x y) dy + n( x) n noise
The role of the OTF (or MTF)
3D Information transfer In analogy to the twodimensional image formation, we can determine a 3D Point spread function (PSF) and a 3D Optical Transfer function (OTF). PSF z OTF kz z=0 z=2µm
In a 3D object we have cross-talk between in- and outof-focus parts In-focus part Out-of-focus part
Result is a blurred image with substantial background intensity
Reduce out-of-focus information by inserting a pinhole Illumination / exitation pinhole emission pinhole confocal planes
Result: much sharper pictures non-confocal = wide-field confocal
In practice, confocal microscopes are point scanners PMT replaces the CCD camera Laser replaces the arc lamp
Thick sample imaging
Image formation in the confocal microscope Again the image is formed by a convolution, but the confocal PSF is smaller and has no butterfly wings. z kz Widefield PSF confocal The optical transfer function has an ellipsoidal shape and has no discontinuity in the middle- optical sectioning confocal OTF
Confocal vs widefield microscope sharp optical sectioning point-scanning method (slow) majority of returned photons not detected wait for a long time to get robust signal even slower Photodetector noise gets critical (weak SNR) Photodamage on sample nice additional features: use programmability of laser scans for bleaching experiments for selective point measurements in small volumes (spectroscopy, fluorescence correlation spectroscopy)
Multi-photon microscopy
Fluorescence fundamentals
2 photon microscopy 10ns 100fs lens Pulsed lasers (typically Ti:Saph ) and tight focusing increase the photon flux. Linescan using confocal and 2 photon microscopy Because of the extremly high photon density at the focal point, it is possible that two photons interact simultaneously with a fluorophore. No bleaching in the out-offocus planes, but increased photo-bleaching in the focal plane (~10faster)!
The multi-photon microscope (in comparison to conventional and confocal microscopy)
The major advantage is the ability to reduce the influence of light scattering in the sample Scattering of emitted rays Capture of scattered, emitted rays Scattering of excitation rays Less scattering of excitation rays (long wavelength)
Demonstration of 2-photon performance on a pollen grain 20 µm
Major advantages (and usefulness) Imaging of scattering samples Deep sections and whole tissue imaging Maximal use of light Shorter exposure times and levels Low photobleaching outside the focal volume Long observation possible Low photo-toxicity
Summary: multiphoton microscopy Thick section imaging Long duration live cell microscopy Lower resolution compared to confocal Long wavelength excitation Thermal damage from chromophores that absorb in the IR spectrum Dependent on fluorescence Expensive (requires a pulsed laser setup)
Fast (camera-based) Inherent sectioning capability (like a confocal), without «throwing» light away Rotation of the sample: uniform imaging (resolution) like tomography Minimized bleaching/photo-toxicity (only the interesting plane is illuminated) Selective plane illumination microscopy Re-discovering a 100 years old idea
Numerous SPIM versions exist
Long term SPIM imaging Tomancak lab
SPIM limitations Sample size (thickness) Sample mounting Aberrations Data amount
The triangle of compromises Signal/Noise ratio (image quality) Image resolution Imaging speed
Deconvolution * = Image is formed by convolution of the 3D Object with the PSF. Can this opertation be inverted?
Deconvolution microscopy: the alternative for rapid 3D imaging modeled PSF measured image unknown object distribution measured PSF unknown PSF unknown noise
The mathematical challenge gˆ( y) = bestmatch i( x), A simple idea: g( y) h( x y) d y Two practical difficulties: 1.) H(ν) is not always positive (bandpass and aberrations) 2.) Noise in I(ν) gets amplified by division by small H(ν)
Deconvolution The inverse operation of the convolution, the deconvolution, is the division of the image spectrum with the OTF. Division by nearly zero and zero is not such a good idea. No information has been physically transfered outside of the support of the OTF (nonzero region), so no information can be reconstructed. Still people try it and corresponding software has become available. Deconvolution techniques: 2D Methods:Deblurring: simply subtracts estimate of out of focus light 3D-Methods: Image Restoration, tries to reassign out of focus light to its source
One possible approach: iterative deconvolution NO Close? YES
Widefield, deblurring, full deconvolution
Widefield, deblurring, restoration XLK2 Cell Exp: 0.5 s Lens: 100x/1.4 Restored Unprocessed Nearest Neighbor Both deblurring and restoration improve contrast SNR significantly lower for deblurred image Deblurring results in loss of pixel intensity Restoration results in gain of pixel intensity
Microtubules in Toxoplasma gondii in the WF Microscope Raw Data Decon d Ke Hu David Roos John Murray Jason Swedlow 2001
Conclusion: Confocal vs. deconvolution microscopy Confocal is the optimal 2D microscope! Deconvolution microscopy is the faster technique in 3D (in acquisition not in data analysis) Where affordable: combine confocal and deconvolution microscopy for optimal 3D imaging