Granularity features

Description

Granularity features measure the size distribution of texture elements (granules) at multiple scales. these features reveal the characteristic scale of intensity variations within objects.

Spectrum approach

Granularity is calculated across a spectrum of scales (1 to 16). in lay terms, granularity calculates how much of an image remains after “eroding” the image over multiple iterations or spectrum of scales. at each scale, the granularity value reflects the amount of texture detail present at that specific size. the feature’s value can be read as the % of the original image signal retained at that scale.

Features extracted

Feature	description
GRANULARITY.1	granularity at scale 1 (finest scale)
GRANULARITY.2	granularity at scale 2
…	…
GRANULARITY.16	granularity at scale 16 (coarsest scale)

Interpretation

High granularity at small scales: fine-grained texture details
High granularity at large scales: coarse regional intensity variations
Granularity profile: overall texture scale characteristics

High granularity at small scales (1-3):

Indicates many small, punctate structures

Examples: individual vesicles, small mitochondria, RNA granules

High granularity at medium scales (4-8):

Indicates larger organized structures

Examples: mitochondrial networks, endoplasmic reticulum sheets

High granularity at large scales (9-16):

Indicates very coarse, chunky texture

Examples: large organelle aggregates, nuclear condensation

Smooth profile (low across all scales):

Indicates uniform, homogeneous intensity

Examples: diffuse cytoplasmic proteins, uniform nuclear staining

As an example, if a drug causes mitochondria to fragment, you’d see:

Increased GRANULARITY.1-4 (more small pieces)
Decreased GRANULARITY.8-12 (fewer large networks)

Applications

Granularity features are useful for:

Identifying cellular texture scale patterns
Detecting subcellular compartment granularity
Characterizing organelle size distributions
Quantifying spatial heterogeneity