Types of analysis

CALMaR analyses answer different questions about the same lesion. Their outputs should not be collapsed into one undifferentiated notion of “brain damage.”

Analysis	Main question	Key dependency
Lesion description	How large is the mask and where is it?	Segmentation quality
Atlas overlap	Which labelled regions directly overlap it?	Registration, atlas choice
Tract overlap	Which reference pathways intersect it?	Tract definition and space
Disconnectome	Which structural connections may be disrupted?	Normative or measured connectivity model
Lesion-network mapping	Which functionally connected regions may be affected remotely?	Normative functional-connectivity data
Meta-analytic decoding	Which research terms are associated with the affected map?	Literature database and decoding method
Lesion-symptom analysis	Which lesion features are statistically associated with behaviour?	Cohort, outcome, covariates, model
Knowledge-base matching	What published findings relate to measured features?	Evidence schema and review quality

Direct measurements and derived inferences

A useful reporting hierarchy is:

1. Observed or segmented: the image and lesion mask.
2. Spatially derived: overlap calculated after documented transforms.
3. Normatively inferred: disconnection or connectivity estimated from another cohort.
4. Literature associated: functions or outcomes linked through published group findings.
5. Clinically interpreted: meaning considered alongside assessment, history, goals, and trajectory.

Each step can be useful. Each step must preserve its provenance and uncertainty.

Analyses not models

Many CALMaR operations are deterministic calculations or evidence lookups rather than newly trained prediction models. Atlas overlap, for example, calculates a geometric intersection. The fact that a later report interprets that intersection does not turn the overlap computation into a learned model.