AHMAD ALMAHOZI

8/30/2025

(gpt) Guide to Casali 2013: (1) General summary

————-

A Theoretically Based Index of Consciousness

Independent of Sensory Processing and Behavior

———————

Casali AG, Gosseries O, Rosanova M, Boly M, Sarasso S, Casali KR, Casarotto S, Bruno MA, Laureys S, Tononi G, Massimini M. A theoretically based index of consciousness independent of sensory processing and behavior. Sci Transl Med. 2013 Aug 14;5(198):198ra105. doi: 10.1126/scitranslmed.3006294. PMID: 23946194.

—————————————

Core question. How can we objectively tell whether a brain is in a conscious state without relying on behavior or sensory processing?

Theory to method. Drawing on Integrated Information Theory (IIT), the paper argues that a conscious brain should generate spatiotemporally integrated yet differentiated activity patterns when perturbed. They operationalize this as Perturbational Complexity Index (PCI): deliver a brief TMS pulse to cortex, record the ensuing EEG response across time and space, then quantify how richly structured that response is using an algorithmic-complexity measure (Lempel–Ziv).

How PCI is computed (high level).

Deliver single-pulse TMS to a cortical site during different brain states (awake, REM/NREM sleep, anesthesia, disorders of consciousness).
Record high-density EEG and reconstruct cortical sources over ~300 ms after the pulse.
At each source and time point, test whether activity is significantly above baseline → binarize into a spatiotemporal “pattern of activations.”
Concatenate that binary matrix into a string and compute Lempel–Ziv complexity (LZc).
Normalize for size/sparsity to get PCI ∈ [0,1].

Main results .

High PCI in healthy wakefulness and REM sleep; low PCI in NREM sleep and general anesthesia.
In disorders of consciousness, low PCI in unresponsive/vegetative states, higher in minimally conscious or locked-in states—tracking the capacity for consciousness rather than motor output.
A single cut-off cleanly separates conscious vs. unconscious conditions across individuals in their sample.

Why it matters. PCI is perturbation-based (causal), content-agnostic (independent of tasks/stimuli), portable across etiologies, and grounded in theory. Clinically, it offers an objective tool to assess residual consciousness when behavior is unreliable.

Theoretical foundation: Integrated Information Theory (IIT)

IIT, introduced by Tononi (2004), says that a system is conscious if it has two properties:

Differentiation (richness): the system can be in a huge variety of distinct states.
Integration (unity): these states are not independent fragments, but influence each other.

Analogy:

A light bulb: 2 states (on/off). Differentiation = very low.
A digital camera with 1 million pixels: many states, but if each pixel works independently, there’s no integration.
A brain: both highly differentiated and highly integrated.

Thus, consciousness = integrated differentiation (high “complexity”).

Why perturbation?

Measuring spontaneous EEG might tell us something, but it’s confounded by:

Background noise
Random fluctuations
Non-conscious processes (like reflex loops)

So Casali et al. use a causal probe:

Deliver a pulse with transcranial magnetic stimulation (TMS).
Observe how the brain’s network reacts.

Idea: If the brain is capable of consciousness, the perturbation will propagate in a complex, integrated-differentiated pattern.

If not, the response will be:

Local only → no integration.
Global but stereotyped → no differentiation.
In both cases, low complexity.

Perturbational Complexity Index (PCI)

Casali et al. invented PCI to capture this idea in a single number:

Perturb the brain (TMS pulse).
Record responses with high-density EEG.
Reconstruct sources → where in the cortex activity originates.
Statistical thresholding → keep only significant activations.
Compress the spatiotemporal pattern using Lempel–Ziv algorithm (a measure of algorithmic complexity).
Normalize for entropy → final PCI between 0 and 1.