Giles, Dominic Matthew; (2023) Representation and Causation in the Focally Injured Human Brain. Doctoral thesis (Ph.D), UCL (University College London).

Text thesis.pdf - Other Access restricted to UCL open access staff until 1 November 2024. Download (146MB)

Abstract

Treatments are prescribed to individuals in pursuit of contemporaneously unobserved outcomes, based on evidence derived from populations with historically observed treatments and outcomes. Optimal treatment selection therefore rests on counterfactual inference of individual potential outcomes. Prescriptive fidelity therefore cannot be evaluated empirically at the individual-level, forcing reliance on lossy, group-level estimates, such as average treatment effects, that presume an implausibly low ceiling on individuation. Where, as in ischaemic stroke, the underlying causal field is wide, reflecting a complex, multifactorial pathological process, the fidelity of such prescriptive inference under prevailing healthcare data regimes is sensitive to the quality of data representation. An optimal representation must be rich enough to capture variations in individual susceptibility while remaining succinct enough for the inference to be adequately supported. In the observational setting, it should moreover be disentangled from non-random treatment allocation so as to isolate the specific causal effect of the intervention. By applying anatomical patterns of ischaemic stroke, in the form of lesion masks and disconnectome distributions, to the task of crafting representations optimised for estimating individualised treatment effects from observational data, their evaluation under observational conditions is enabled when combined with an empirically informed semi-synthetic simulation framework. This uses transcriptomic and receptomic data to model heterogeneous treatment effect sizes and variability, and observable and unobservable confounding treatment allocation biases, with explicit modelling of decoupled response failure and spontaneous recovery. The evaluation of representation–prescription models, flexible enough to capture the observed heterogeneity, found that the richness of the modelled lesion representation is decisive in determining individual-level fidelity, even where freedom from treatment allocation bias cannot be guaranteed. I am compelled to conclude that complex modelling of richly represented data is crucial for individualised prescriptive inference in ischaemic stroke.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item