Fundamentals of Computational Bioimage Analysis Flashcards
Why use digital image analysis?
- Can be automated
- Repeatable
- Reproducible
- Clearly-defined and documented
- Produces numerical data for statistics
What are the components of an image?
- X,Y dimensions (optional Z, time, channels)
- Each pixel records a signal intensity at a given bit-depth (12-bit, 16-bit, 24-bit colour usually used to quantify sizes)
How are images captured?
- Cells are marked with fluorophores for components of interest
- Fluorophores are excited and separate images captured for each channel, timepoint, experimental condition
What are the compromises to be made for each experiment?
Choose between resolution, speed, photo-damage and SNR
What are the questions to be asked before designing an imaging experiment?
- What is the question?
- Can everything be detected?
- What does a false positive/negative look like?
- What needs to be measured?
- What needs to be defined?
Define
SNR
Signal to noise ratio, a measure of the quality of the signal
Define
Segmentation
Splitting an image into regions relevant to the question being asked (cell, nuclei, puncta, tissue regions)
Define
Thresholding
Splitting an image into 2 components based on an intensity value (can be fixed or mathematically derived (Otsu, Tringle))
Define
Filtering
Application of mathematical filters to the image, used to highlight features and improve segmentation
Define
Background subtraction
Removal of background that appears due to out of focus light, mounting/culture media or poor fixation
Removed using top-hat filter or rolling ball method
Define
Local Minima/Maxima
- Find bright or dark spots in an image
- Sensitivity can be adjusted to cope with noise
- Pre-filtering improves accuracy
Define
Watershed
- Allows separation of touching objects
- Can be performed in binary image using distance transforms
- Can be performed with minima/maxima in gradient images
