Key Points

Opening and Checking an Image

Standard image formats (TIFF, PNG) can be loaded as NumPy arrays using skimage. Proprietary microscope formats are best handled by BIOIO to preserve dimensions and metadata. Basic metrics like histograms, shape, and pixel ranges help determine the best analysis strategy. Lookup tables (LUTs) change how data is rendered visually but do not change the underlying pixel values. Lazy loading (via Dask in BIOIO) allows you to explore massive datasets without overwhelming your computer’s memory.

Applying Filters

  • There are many ways of smoothing an image
  • Different methods will perform better in different situations

Thresholding and Segmentation

  • Thresholding, labelling and feature isolation are all forms of segmentation
  • Different thresholding algorithms can perform differently in different situations
  • Quantitative image analysis often requires us to individually label features
  • Processing such as watershed transforms can solve cases where features of interest are very close or stuck together

Measurements

  • Binary multiplication and numpy.where are powerful ways of combining images and masks
  • The numbers assigned during labelling can be used to select and process one feature at a time

Introduction to Napari

  • Napari works with layers, each of which represents an image channel
  • Napari doesn’t always know how to load a multi-channel image from the GUI
  • We can use the Python console to perform custom operations that can’t be done in the GUI
  • Most operations we performed earlier in Python can also be done in Napari, either graphically or in the terminal