Abstract
Early diagnosis and efficient treatment for Alzheimer's Disease (AD) remain elusive. Current efforts
are aimed at a generalized patient profile, overlooking many nuances of individual disease
trajectories. Consequently, the efficacy of a treatment comes with a high uncertainty. The
fundamental issue lies in the limited understanding of the initial causative triggers of the disease and
heterogeneity of AD, resulting in treatments that merely address the current manifestations rather
than targeting the root of the problem.
To address this, individual disease trajectories need to be obtained, allowing analysis of causative
triggers and prescription of treatment at a per individual level. However, health records are typically
only obtained after clinical onset, which occurs after extensive neurodegeneration at the end of AD
pathology.
As a solution we propose an in silico method to counterfactually rejuvenate individual health records,
obtained after clinical onset, to their pre-disease state. We leverage causal theory as it provides a
basis to rejuvenate health records. We develop causal-capable neural networks incorporating
multimodal data to rejuvenate, based on available brain-ageing-neural-network and causal-neuralnetwork.
Obtained individual trajectories allow personalized analysis of disease triggers and effective
personalized treatment. Longitudinal database of health trajectories covering whole AD pathology will
be created, allowing discovery of early AD biomarkers.
Onderzoeker(s)
Onderzoeksgroep(en)
Project type(s)