Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Week-long microscopy course Class 2 http://microscopy.duke.edu/ [email protected] The problem with standard microscopes and biology Sectioning Optical sectioning of thick samples 3D reconstruction The confocal principle How a laser scanning confocal microscope works http://www.olympusfluoview.com/theory/index.html What a confocal looks like Scanhead A microscope (inverted or upright) Lasers and electronics Probably on an air table Computer Confocals have lots of adjustability LASERs are used for excitation Ideal for point scanning: •Narrow collimated beam, low divergence •Powerful http://www.olympusfluoview.com/theory/laserintro.html Many lasers available . . . Gas lasers Diode lasers (Kr/Ar 488 568 647) 405 Ar/Ar 458 488 488 HeNe 543 594 633 514 561 635 (Color coding refers to the color of the fluorophore for which the laser line is most commonly used) Adjusting the laser power Relatively fixed output Rapid control 0-100% T Laser AOTF Laser AOTF l selection and intensity control Fluorophore saturation Bleaching is proportionately worse here Emission intensity Image will suffer loss of contrast and quantification issues Excitation intensity Widefield is normally in the linear range, the concentrated laser spot in confocal may not be. Start low and increase to the minimum necessary Photomultiplier tubes (PMTs) are used as detectors • • • • • • Fast (good for scanning) Large collection area Good SNR Very large dynamic range (with gain&offset adjustment) Adequate dynamic range at a single gain&offset position QE <30% (not as good as CCD) Gain and offset adjustments 100 msec 200 msec 400 msec 600 msec 800 msec 1000 msec 2000 msec Camera: Confocal: Increasing gain (voltage on PMT) Offset to set the background to black Optimal gain and offset Confocals have special display modes to highlight saturated and 0 intensity pixels Should you have no, a few or many saturated and zero intensity pixels? Scanning mirrors The relative position of the two mirrors can point the spot anywhere in the field http://www.olympusconfocal.com/theory/confocalscanningsystems.html Scan speed Fast scans • Fast processes • Useful for focusing and adjustments – eg 1 fps Slower scans • More light gathered • Better images • More damage Averaging Scanning each pixel multiple times and averaging improves the noisy signal . . . Averaging: how much do you need? 1 2 8 4 16 Decreasing rate of improvement, empirically determine a good balance between final SNR and time/damage in acquisition Quiz B Scan area: zoom The area swept by the galvo mirrors can be adjusted . . . Is this meaningful zoom or just digital zoom? . . . Scan area and number of pixels Any particular frame can have different numbers of pixels. . . How do I set these two things? How many pixels do I need in my image for the best resolution? Same area: 225 mm across 225 mm/512 = 440 nm 225 mm/1024 = 220 nm The Leica SP5 goes up to 8K by 8K, shall we have 64 Mpx for every scan? What was a pixel? Part 1. • • • • • Shall we have as many px as possible? How big should a pixel be? Are trade-offs involved? Are pixels real? What is a grey value? How many pixels do I need in my image for the best resolution? Nyquist sampling theorem: Sample at twice the resolution Resolution = 0.61 l / NA Increasing number of pixels per area Signal under-sampled Not capturing all the resolution of the system Signal Well sampled Just right, The Nyquist rate Signal Over-sampled Not gaining any more resolution, more bleaching waste of time and disk space Do I really need to listen to Nyquist? • You might not always be seeking the best resolution • You might need to under-sample for speed, phototoxicity . . . • Over-sampling is effectively averaging But optimal sampling is important and beneficial in many cases Lateral resolution in confocal: Theory At best, the lateral resolution of a confocal is better than widefield by a factor of √2. wf confocal Essentially the effective airy disk is the product of illumination and detection disks so has steeper sides Lateral resolution in confocal: Reality But to see this benefit the pinhole needs to be at <0.25 AU. rxy Relative intensity 0.61 l √2 NA 10% ~ 0.61 l NA ~ 0.61 l NA 100% Up to 1000% Quiz C Multi-channel confocals Most confocals have several laser lines and more than one PMT Why are fluorescence scopes generally not built like this? Simultaneous or sequential acquisition Blue Green Red Faster Blue then Green then Red Less bleedthrough Bleedthrough DAPI AF488 • Down the spectrum (1st law of thermodynamics) • Worse when intensities are unbalanced Line vs frame switching Blue line1 then Green line1 then Red line1 Blue line2 then Green line2 then Red line2 The AOTFs switch the laser light on and off very rapidly Good to see all the channels appearing together Blue image then Green image then Red image Good when something physical happens between colours (eg change dichroic) Use this when you have more channels than PMTs 3D acquisition Confocals are good for acquiring stacks of images because of the optical sectioning ability Lateral and axial resolution XY rxy ~ 0.61 l / NA YZ rz l / NA2 Resolution is always worse in Z than XY Optical section thickness 10x/0.3 20x/0.45 Optical section (micron) Smaller with higher NA 63x/1.2 40x/1.3 63x/1.4 100x/1.4 Larger with more open PH Pinhole size (AU) Sampling in the z-axis The same principle as in XY Some regions not imaged Covered Covered and well sampled Imaging big things Macroscope | Stereoscope Macro/stereoscope imaging modes Stitching/tiling/montage/mosaic Two photon excitation Single photon Two-photon Excitation is limited to a small focal volume where photons are most concentrated Differences to a single photon confocal Can use all the light, no pinhole needed 2-photon advantages Main advantage: Imaging thicker specimens The longer wavelength excitation penetrates further into the sample (less scatter and absorption in IR). The scattered excitation light doesn’t cause background fluorescence. The excitation is also not attenuated by fluorophore absorption above the plane of focus Emission advantage Because we are able to image all the light (no pinhole) this is less affected by scatter With pinhole No pinhole The NDD is closer and more efficient for a scattered beam path (which is hard to move efficiently through several lenses) 1P 2P Pretty pictures from the Olympus FV1000 MPE brochure A Pulsed laser is required for MPE For efficient MPE we need photon concentration in space and time . . . http://www.olympusmicro.com/primer/techniques/fluorescence/multiphoton/multiphotonintro.html A pulsed laser Chameleon Ultra II Femtosecond pulsed laser 3-4 W of power at peak Tunable 680-1080 nm @40 nm/sec This gives flexibility, since these laser cost about $200,000 each its not normally practical to have several per machine like with 1 photon Fluorophores for 2-photon Some fluorophores are pretty different between 1P and 2P Excitation spectra not exactly double 1photon (tends to be broader and blue shifted) Quantum dots can be excited at nearly any l Fluorescent proteins are pretty good Photobleaching and toxicity Photodamage in 2P is confined to the thin layer being imaged Power (photons/mm2 Wide-field X 1 Photon 105X 2 Photon (average) 106X 2 Photon (peak) 1011X The bleaching in that small area is probably worse than 1P (some photodamage is highly nonlinear) In general, for thick samples 2P has an advantage over 1P For thin samples, 2P is often worse than 1P SCALE reagent • • • • 4 M urea 10% (wt/vol) Glycerol 0.1% (wt/vol) Triton X-100 pH of 7.7 Refractive index of 1.387 and 486 nm Soak your sample in it for a couple of weeks . . . http://www.nature.com/neuro/journal/v14/n11/full/nn.2928.html It clears fixed samples by removing refractive index changes Doesn’t destroy fluorescence Preserves tissue structure And allows . . . 2-photon vs. 1-photon •Improved SNR in deep-tissue imaging (especially with NDDs) •The IR 2PE is less phototoxic in many cases, especially for UV dyes. •Photobleaching/damage restricted to plane being imaged. Photobleaching or uncaging is possible with fine z-axis resolution • XY resolution is slightly less good •Multi-channel acquisition is harder (excitation cross-sections are normally much broader) and limited in excitation l •No control of optical section thickness • Lasers are expensive and require exact alignment and can produce heating and other damage Don’t use a 2-photon system unless you need the advantages Fast confocals Other ways of doing something like this Without these problems . . . • • • What can we do about this? • • • • • • The spinning disk principle The spinning disk principle Detector comparison Photodetectors Photobleaching advantages t t m How does a spinning disk work? Sectioning in spinning disks Ideal pinhole diameter = 0.5 l M/NA • 100x/1.4 = 20 mm • 20x/0.5 = 11 mm What works well on a spinning disk Living things that need sectioning . . . • Things which match the high magnification, high NA optimizations (eg subcellular imaging) • Photosensitive samples • Fast imaging Quiz D Deconvolution A mathematical post-acquisition processing of images to reduce the blur from out of focus light. This can increase the signal to noise ratio and resolution of the image. The Point-Spread Function and image formation XZ XY PSF Image The image is the sum of all blurred point images Convolution An image is a convolution of the object: Object x PSF = Image XY XZ The aim of De-Convolution Image(s) Knowledge of imaging psf Underlying object The PSF can be measured or predicted Two types of deconvolution 1. Deblurring/nearest neighbor/2D deconvolution – Estimates the blur from other focal planes and removes it – Sharpens the image but is non-quantitative – Very fast, real time XZ Blur the image from here and remove it from the images above and below Two types of deconvolution 2. Restoration/3D deconvolution o Iterative reassignment of photons based on modeled convolution o Works in 3D (ie considers all the data together) o More computational intensive, takes a few seconds to minutes o Conservative and quantitative o Better able to cope with low SNRs Move this photon from the blur to the object Deblurring v restoration A Workingperson’s Guide to Deconvolution in Light Microscopy The computational process Original image Estimated Object psf convolve Predicted Image Constraints n iterations The psf is improved in blind deconvolution Photon reassignment Compare to original image Stopping criteria: If the same, the object perfectly accounts for the image Finished What actually improves during deconvolution? • Photon reassignment: • Noise Blur Structure . . . So higher SNR Which modalities does it work with? All of them but the relative benefits are different for the different modalities Widefield PSF calculated for 60x/1.4 NA objective Green fluor = 2 mm Spinning disk Confocal (1 AU) Limited in widefield for thick samples Too many PSFs to draw Image: Sharp image +dimmer blur Overwhelmed by blur Summary: What deconvolution is/ is not Post-processing that works for all (3D) imaging modalities Fairly computational intensive to calculate properly It needs good images - Does NOT allow you to take awful images and magically transform them! Good for live cell imaging: gentle but slightly noisy images + deconvolution = good images with less phototoxicity It doesn’t replace confocal. Anything >30 mm and the blur becomes too much for processing and you need a spatial filter. TIRF Total Internal Reflection Some reflected, some refracted Evanescent wave Evanescent wave Exponential decay of intensity Iz = I0e-bz Evanescent wave Evanescent wave Plane of excitation ~100 nm thick Two ways of generating and imaging TIRF Prism-based A bit more to align Better SNR, lower background Slight constraint on imaging objective Sample access is difficult in some setups Have to build your own Objective-based More convenient Needs >1.45 NA objective SNR still very good Components of a TIRF system TIRF angle autoalignment optics round the back Opaque incubator Fiber-coupled lasers EMCCD Inverted microscope with a special TIRF objective Widefield vs TIRF Myosin Actin What is TIRF good for? Anything at the edge of the cell/tissue • • • • Exocytosis/endocytosis Vesicle dynamics Cytoskeletal activity at the membrane – Focal adhesions Signalling in the membrane – translocation Relatively distinct subset of samples gain from TIRF imaging . . . Tools available for you Choose wisely Further information Review articles about confocal theory http://www.olympusfluoview.com/theory/index.html Java tutorials to play with http://www.olympusfluoview.com/java/index.html SVI wiki on 3D Microscopy and deconvolution http://support.svi.nl/wiki/ Deconvolution review articles http://zeiss-campus.magnet.fsu.edu/referencelibrary/deconvolution.html Multiphoton articles/tutorials http://micro.magnet.fsu.edu/primer/techniques/fluorescence/multiphoton/multiphotonhome.html Reviews on multiphoton imaging http://zeiss-campus.magnet.fsu.edu/referencelibrary/multiphoton.html TIRF http://www.microscopyu.com/articles/fluorescence/tirf/tirfintro.html SPIM http://dev.biologists.org/content/136/12/1963 Spinning disks http://zeiss-campus.magnet.fsu.edu/articles/spinningdisk/introduction.html