Confocal and Deconvolution Microscopy Notes

Confocal and Deconvolution Microscopy

Confocal Imaging: Basic Principles

  • Confocal imaging employs a light source, typically a laser, which is deflected by a dichroic mirror to provide coaxial illumination.
  • Modern confocal designs utilize collimated laser light and an infinity optical path.
  • The collimated beam forms a single 3D Point Spread Function (PSF) within the sample volume, eliminating the need for a field aperture as used in conventional fluorescence or bright field microscopy.
  • Instead of a camera, a single detector measures the brightness of emitted light.

Key Components of Confocal Microscopes

Light Sources (Older Confocal Light Path - Zeiss LSM 510 meta):
  • Blue laser diode (405 nm)
  • Ar ion laser (458, 477, 488, 514 nm)
  • Green DPSS laser (561 nm)
  • HeNe gas laser (633 nm)
  • Multiple lasers can be used simultaneously or sequentially.
  • Light enters the path via optical fibers.
  • Beam-splitters and dichroic mirrors separate multiple light paths to multiple detectors, enabling simultaneous or sequential illumination of multiple fluorophores.
Pinholes (Confocal Apertures):
  • Pinholes eliminate out-of-focus light.
  • There is one pinhole per detector.
  • Pinholes are individually adjustable.
  • Typically adjusted to match the diffraction image for each wavelength (longer wavelength = larger aperture).
  • The unit of measure is an "Airy unit" (A.U.) pinhole.
Detectors:
  • PMTs (photomultiplier tubes)
  • Spectral detector (e.g., “metadetector”) – a spectrometer that allows multiple wavelengths to be analyzed.
  • Photon-avalanche detectors
  • Lambda (λ\lambda) scan detector array can:
    • Detect different channels (=wavelength intervals) simultaneously.
    • Utilize dichroic filters (on motorized wheel) or spectral ‘sliders’ to allow continuous variation of the wavelength range captured (most modern systems).
Lambda (λ) Scan:
  • Meta-detector array can detect up to 32 different channels (=wavelength intervals) simultaneously.
  • Allows spectral properties of emitted fluorescence to be quantified.
  • Each image represents the emission intensity distribution within a wavelength interval.
Spectral Unmixing:
  • Lambda (λ\lambda) scan facilitates spectral ‘unmixing.’
  • Spectral emission curves can be obtained for each pixel or region of interest.
  • Two or more fluorophores can be ’unmixed’ by comparison with reference spectra.
Dichroic Mirrors:
  • Dichroic mirrors at the detector are inefficient.
  • They pass desired wavelengths (to match fluorophore emission spectrum) and reflect other wavelengths.
  • Detectors are typically arranged in sequence so that reflected light can be scavenged (e.g., during simultaneous excitation with 2 or more lasers).
  • Reflected light can be lost due to limits on the number/specificity of filters included in the system.

New Generation Confocal Laser Scanning Microscopy (CLSM)

  • Solid-state lasers (approximately 1/4 the cost of older gas lasers).
  • Replaces dichroic mirrors with a spectrophotometer mechanism for more efficient light collection.
  • Typically uses detectors (e.g., GaAsP) with higher quantal efficiency.
  • Detection wavelength band can be adjusted in very small increments (~1 nm).
  • Leica TCS SP8 is an example.
  • Allows more of the emitted signal to be detected, reducing scan time, photobleaching, and other issues.
  • On the Leica SP8 system, there are 4 visible light lasers + 1 “UV” (actually 405nm).
  • Light (all colors) passes a single pinhole before a prism splits the wavelengths spatially.
  • Broadband mirrors are combined with moveable knife-like spectral ‘sliders’ to split the spectrum up between detectors.

Photomultiplier Tubes (PMTs)

  • Highly sensitive detectors used in confocal imaging.
  • PMTs sample from one point; hence, the image is built up sequentially, one pixel at a time.
  • There is a drive for higher sensitivity to reduce required ”pixel dwell” time and increase scan speed.
  • PMTs have very high sequential readout speed (pixel dwell in the nanosecond to microsecond range), low (or no) dark noise, and good quantal efficiency.

New Technologies (GaAsP PMTs & Hybrid Detectors)

  • Improve sensitivity further.
  • Hybrid detectors have essentially no dark noise.
  • More expensive.
  • Can be damaged by excessive light.
  • Less sensitive at very long wavelengths.

Confocal Laser Scanning Microscope Operation

  • The focal plane is shifted by movement of the stage up or down in precise steps to provide information in the 3rd dimension (Z).
  • The image is built up by scanning the light source (a laser) in a sequence of scan lines that move the focused spot across the specimen anywhere within a large volume. (XYZ)
  • A detector collects the emitted light (a time series) and bins it into chunks to create virtual pixels (or 3-dimensional pixels – ‘voxels’).
  • Individual ‘pixels’ at each X, Y location are acquired by integrating the signal from the detector (a time series) at each sequential location for a given ‘pixel dwell time’ (e.g., 20µs).
  • Information in the 3rd dimension can be obtained by shifting the specimen (stage) towards or away from the objective lens and repeating the scan (X, Y, Z).
  • 1 voxel = 3-dimensional pixel
  • Note the trade-off between the number of scan lines, the number of pixel subdivisions, and the dwell time (e.g., 20µs). If we want more pixels, the scan duration increases or dwell times become very short.

Working with Volumes

  • A “z-stack” of images of consecutive confocal sections is collected.
  • “projection” (maximum intensity projection or MIP) of all confocal planes shows all structures in focus simultaneously.
  • 3D surface rendering.

3D Analysis

  • Analysis involving two or more Fluorophores
  • Orthogonal views make it possible to investigate interrelationships of labeled structures (XZ, XY, YZ).

CLSM: Pros and Cons

Advantages over wide-field fluorescence:
  • Better lateral (x/y) resolution.
  • Better depth (z) resolution (ideal for 3D reconstruction of large volumes).
  • Better visibility of fine details.
  • Better discrimination for multiple fluorophores.
Disadvantages over wide-field fluorescence:
  • Sequential scanning of each point can be very slow.
  • High illumination intensity can bleach fluorescence.
  • Very high cost (typically >3x high-end fluorescence microscope).

Deconvolution

Deconvolution + Confocal -> super-resolution in 3D

  • (f∗h)(f * h) is, technically, the integral of the product of f and h.
  • Deconvolution in microscopy attempts to estimate f, given a model for g and on the assumption that the image represents (f∗h)+noise(f * h) + noise.
  • Deconvolution is a general term: there are a large family of different methods operating in 2D and 3D.
  • It removes out-of-focus blur or re-assigns it back to its point of origin by collecting a stack of images (‘XYZ’ series or ‘Z-stack’) from different focal planes, similarly to histological sectioning.

Iterative Deconvolution

  • Starts with an initial guess of the real object (i.e., what the real images should be like).

  • Convolves (blurs) this estimated object using the PSF.

  • Compares the ”blurred images” to the raw observed images and computes an ”error criterion.”

  • Uses this comparison to improve the estimate of the object.

  • Repeats the process until the error criterion is minimized.

  • The estimation of the object at this stage is now the ”restored image.”

  • Equation: f<em>k+1=f</em>k+(g−fk⊗h)⋅h∣h∣2f<em>{k+1} = f</em>k + (g - f_k \otimes h) \cdot \frac{h}{ | h |^2}

Practical Considerations

  • Sampling interval

  • Nyquist theorem: In order to fully describe objects, they must be sampled at a rate of twice the smallest resolvable object
    XY D=0.61λ\lambda/NA Z D=2n*λ\lambda/(NA)2^2

  • Need to match our camera or detector to the image size (we will come back to this point in the next lecture)

  • In confocal imaging, we have no camera… Pixel size is set by the scan field we specify (i.e., we can make the pixels as large or small as we want

  • For best resolution, we want to use ”2x oversampling”: i.e., we make the optical slices (Z) and XY pixel size ½ the size suggested by Rayleigh’s criterion

  • SLOW!

More Practical Considerations

Artefacts
  • Interpret with care!
  • Deconvolution algorithms can easily lead to artefacts in the images
Sources of error
  • The quality of raw data is very important
  • Accurately determined PSF that is valid for the imaging conditions
  • Avoid spherical aberration
  • Reduce noise
  • Compensate for photobleaching
  • Avoid lamp flickering
  • Avoid ‘saturation’ of the detector!