Scalable Biomarkers and Generative Digital Twins for Personalized Neurostimulation in Depression

点击翻译按钮获取中文摘要

Project Summary/Abstract More than 100 million people in the United States currently show signs of clinical depression, approximately three times more than before the onset of the COVID-19 crisis. Currently, the main treatment options for such depressed individuals include pharmacological and psychological interventions, the acute and long-term effectiveness of which are significantly limited: up to one-third of patients develop treatment-resistant depression. Noninvasive neurostimulation therapies such as repetitive Transcranial Magnetic Stimulation (rTMS)–where a magnetic coil placed over the cortex is used to focally stimulate the brain–have recently emerged as promising low-risk interventions for treatment-resistant depression. However, the mechanisms and appropriate parameters for this treatment remain poorly understood. In the best cases, rTMS can have dramatic effects, changing the course of a patient's life in hours. In many cases, however, it has little to no measurable effect. This raises the obvious question: why do current rTMS protocols work well for some individuals but not for others? Could we adapt protocols to work well for everyone, potentially providing reliable personalized treatment or even a lasting cure? A growing literature suggests this is possible if we learn to tailor treatment to individual differences in human neurophysiology. Here we propose an innovative and unique approach towards precision psychiatric neurostimulation: personalized modeling of treatment using “Generative Digital Twins”. To affordably build and scale Generative Digital Twins, we propose combining high density electroencephalography (HD-EEG)–which is non-invasive, inexpensive, and easily deployable–to measure longitudinal changes in brain connectivity during an accelerated rTMS treatment protocol for depression. Then, using controllable generative neural networks that allow detailed predictive simulations of individual response trajectories given rTMS treatment, we can begin to predict outcomes prior to treatment, understand individual responses, and personalize treatment parameters.