Research
Neuropsychological task development for epilepsy populations
If stimulation changes something about how a person thinks, a task has to be sensitive enough to show it. Most standard neuropsychological tests were not built for that job. They were built to detect impairment once, in a quiet room, in a session that can take an hour — not to detect a shift, repeatedly, on a hospital ward where the session may end whenever the clinical team needs the patient.
This strand of the work designs tasks for the situation as it actually is. They are short enough to repeat within a session, structured so that a stimulation block and a sham block can be compared without a practice effect swamping the difference, and built to be interrupted and resumed. They cover memory, language, and executive function, because those are the functions the structures being stimulated are implicated in.
The validation question comes first: before a task can be used to test whether stimulation did anything, it has to be shown that the task measures what it claims to, in this population, at this length.
Methods and technical detail
Design constraints
Repeatable within a session without a practice effect large enough to mask the contrast of interest; parallel forms where repetition is unavoidable; interruptible and resumable; administrable at a bedside with no dedicated equipment; short enough to fit the windows a monitoring unit’s schedule leaves open.
Domains
Memory, language, and executive function, selected to line up with the structures the stimulation protocols target rather than to produce a general cognitive profile.
Validation
Psychometric checks on the parallel forms, practice-effect estimation across repeated administrations, and comparison against the clinical neuropsychological battery these patients already receive as part of their surgical workup.
The Electronic Letter Alternation Task
ELAT is an executive-function measure built to distinguish frontal from temporal lobe cognitive deficits during epilepsy surgery evaluation — a distinction the standard battery makes poorly. It is one of three tasks developed independently in this strand, alongside TRT and CDT, and has been presented at conference.
Stack
Task presentation and scoring in Python, with response logging on the same timebase as the electrophysiology so that behavioural and neural effects can be related trial by trial rather than block by block. Statistical modelling in R and SPSS. One task uses an LLM-based sentiment model to quantify the valence and arousal of its word stimuli, which is how the socioemotional work gets a continuous measure out of material that would otherwise need hand-rating.
Outputs
- Tasks Three assessment tasks built independently — ELAT, TRT, and CDT
- Presentations ELAT, chronoception in temporal lobe dysfunction, and epileptic network laterality