ECSE6963 Biological Image Analysis

1
ECSE-6963Biological Image Analysis

• Lecture 4
• Basics of Biological Microscopy
• Badri Roysam
• Rensselaer Polytechnic Institute, Troy, New York
  12180.

Center for Sub-Surface Imaging Sensing
2
Recap

• Basics of biology and medicine
• Important to learn some of the terminology know
  the driving issues
• The Internet has sufficient reference material
  for our purposes
• Molecular Biology (genomics, proteomics) at the
  center of action
• Medical business driven by morbidity and
  mortality statistics
• Biomedical Imaging is key to RD, and clinical
  practice
• Plenty of opportunity for image analysis

3
Basics of Biological Microscopy
Specimen Preparation
Microscopy
Image Capture
Image Analysis

• Why bother?
• Successful image analysis rarely happens at first
  try
• Great looking images are often terrible for image
  analysis
• Sub-visual artifacts from lossy compression, use
  of indexed color instead of true color,
  background noise that is hidden by color maps,.
• Knowing basic terms and issues can help the
  biologist and the computer scientist work out the
  best combination of specimen preparation, image
  capture, and image analysis steps.

4
An Example

• A different color scale (heated object
  colorscale) reveals background noise better

5
Example - Continued

• 1 micrometer spacing between optical slices
  note the sudden change in object shape

6
Example - Continued

• 0.5 micrometer spacing between slices much
  better at preserving structural continuity
  easier to do 3-D image analysis

7
Example Bleed-through
Extracted Transcription Foci (looks clean)
8
Bleed through
Clean-looking Transcription Foci
After Image Re-scaling!
9
What we are after

• Structure of specimens
• The micro-anatomy of biological objects, as
  revealed by a variety of physical properties
  (transmissivity, reflectivity, phase,)
• Function of parts of specimens
• The location activity of specific substances
  within cells and tissue
• Combined structural and functional imaging tells
  us where the function is happening relative to
  the specimen anatomy
• Most powerful and useful

10
Imaging Methods

• Light Microscopy
• Transmitted-light (brightfield) microscopy
• Confocal microscopy
• Phase contrast microscopy
• DIC microscopy
• Fluorescence Microscopy
• The most useful functional imaging tool

11

• Standard Light Microscope (Olympus BH2)

Objective lens The most important part
12
Numerical Aperture (N.A.)
n refractive index
13
Airy Pattern Resolution
Resolution The ability of a microscope to allow
one to distinguish small, closely situated objects
To reveal finer details, we must use smaller
wavelengths of light, and a higher numerical
aperture (N.A.)
14
Resolution Wavelength
mid-spectrum wavelength 550nm
To reveal finer details, we must use smaller
wavelengths of light, and a higher numerical
aperture (N.A.)
15
Fluorescence Microscopy
Excitation Light
Emitted Light
Fluorescence ? Intensity2

• Main Advantages
• Specificity Fluorescent substances are usually
  very specific about excitation emission
  wavelengths, and only fluoresce when the
  excitation is ON.
• We attach fluorescent molecules (fluorochromes)
  to the molecules we want to study.
• Sensitivity Using highly sensitive detectors and
  carefully chosen filters, one can image as few as
  50 molecules per square micrometer.

16
Fluorescence Microscopy
Generally, preferable to illuminate from above
(epi-illumination) rather than from below
(transmitted fluorescence illumination).
17
Practical Issues

• Need to choose fluorophore molecule carefully
• Smaller molecules penetrate specimen better
• Need to tradeoff image brightness with specimen
  damage
• Brightness of image
• E.g., other things being equal, a 40X objective
  with an N.A. of 1.0 will yield images more than
  five times brighter than a 40X objective with a
  numerical aperture of 0.65.
• Electronic sensors give much higher sensitivity
  compared to film
• Photon noise usually a problem
• Fading - These are conditions that may affect the
  re-radiation of light and thus reduce the
  intensity of fluorescence.
• Known as photobleaching, and quenching
• The fluorophore gets tired and damaged under
  intense excitation light

18
Electronic Detectors

• PMT photo multiplier tubes
• CCD charge coupled device

19
The Airy Pattern in 3-D
Also known as the point-spread function
20
Achieving Fine Axial Resolution

• Finer Physical Sectioning
• Limited in scope
• Distorts specimen
• Destructive 3-D imaging possible
• Optical Sectioning
• Method 1 (hardware method) Build a microscope
  that only extracts the light from the best-focus
  plane, and rejects light from above and from
  below
• Method 2 (software method) Develop algorithms
  that attempt to eliminate the light from above
  and below based on a mathematical model of the
  point spread function
• Method 3 combine 1 and 2.

21
Confocal Microscopy (hardware method)
22
Effect of Pinhole
23
Confocal Microscopes

• The diagram on the previous page shows how to
  image one point in the specimen
• The specimen is stepped along x, y and z
  directions to capture a full 3-D digital image
• Elaborate, precise, expensive (about 100K)
  computer-controlled instruments, usually shared
  by many users

24
Fluorescence Confocal Imaging A Great
Combination!
XY
YZ
Alexa Dye Injected Neuron Image Dimensions
512x480x301
XZ
25
Deconvolution The Software Method to Optical
Slicing
Point-Spread Function h(x, y, z)
True Image i(x,y,z)
Observed Image y(x,y,z)
In practice, its complicated by the fact that H()
has zeroes (more later!).
26
Deconvolution Example (software method)
Before
Rat Pyramidal Neuron Stained with HRP
After
27
Confocal Deconvolution Example
Before
After
Top View
Top View
Side View
Side View
Rat CA3 Hippocampal neuron image (Data Courtesy
Dr. James Turner, Wadsworth Center, Albany, New
York)
28
Which Method to Use?

• Widefield Valuable whenever 3-D measurements are
  needed
• Confocal We see from the confocal neuron example
  that the software deconvolution method has the
  greatest impact on axial resolution the lateral
  resolution is not improved much.
• The software method is computationally intensive
  (lots of 3-D Fourier Transforms)
• So, when axial resolution is critical, the
  software method is valuable worth the
  computation.

29
Summary

• Basics of biological microscopy
• Transmission light microscopy
• Fluorescence light microscopy
• Confocal microscopy
• Phase objects
• Phase contrast microscopy
• Differential Interference Contrast Microscopy
• Combined methods
• Multiple fluorophores, in combination with other
  modalities to provide structural and functional
  imaging
• Reference http//micro.magnet.fsu.edu/primer/inde
  x.html

30
Instructor Contact Information

• Badri Roysam
• Professor of Electrical, Computer, Systems
  Engineering
• Office JEC 6046
• Rensselaer Polytechnic Institute
• 110, 8th Street, Troy, New York 12180
• Phone [phone]
• Fax [phone]
• Email roysam_at_ecse.rpi.edu
• Website http//www.rpi.edu/roysab
• NetMeeting ID (for off-campus students)
  128.113.61.80
• Secretary Jeanne Denue, JEC 6049, (518) 276
  6313, denuej_at_ecse.rpi.edu