Prediction of the Cognitive Effects of Electroconvulsive Therapy Via Machine Learning and Neuroimaging

Overview

The study aims to use machine learning to predict the occurrence of episodic and autobiographical memory deficits as well as treatment response following a course of electroconvulsive therapy. Additionally, the neurophysiological correlates of the cognitive effects after a course of ECT will be investigated. Therefore, structural, resting-state and diffusion tensor images will be collected within one week before the first and after the last ECT treatment from severely depressed patients. Standard measures of cognitive function and specifically episodic as well as autobiographical memory will also be collected longitudinally and used for prediction. The study consists of 60 ECT receiving inpatients suffering from major unipolar or bipolar depression, 60 medication-only controls and 60 healthy controls.

Due to the immense disease burden of major depression and unsatisfactory response to standard pharmacological and psychological treatments, the need for treatment alternatives is evident. Electroconvulsive therapy (ECT) remains to be the most efficacious treatment known for treatment-resistant depression. However, although many studies show response rates above 70%, ECT can be considered vastly underused. Reasons contributing to this phenomenon may include stigma, regulatory restrictions, limited medical training, safety and side-effect concerns, or reluctance among professionals to recommend ECT. Most of these reasons have already been refuted or put into perspective by psychological and neuroscientific studies (e.g. ECT causes brain lesions) and most cognitive deficits related to the ECT course seem to fade after several weeks of discontinuation. Still, in terms of the tolerability, memory disturbances remain the most problematic effect of ECT. Besides subjective reports from patients after a course of ECT, experimental studies have also found evidence of episodic and autobiographical memory impiarment attributable to ECT. The origins of these effects are still largely unknown and remain a goal for further research. It has now been shown that structural T1 weighted MR-images can be used to predict the response to a course of ECT via machine learning. Therefore, this study aims to use machine learning to predict the occurrence of episodic and specifically autobiographical memory deficits arising within a course of electroconvulsive therapy based on MR-images collected within one week before the first ECT treatment from severely depressed patients. Additionally, the neurophysiological correlates of the cognitive effects modulated by a course of ECT will be investigated longitudinally through the use of structural, resting-state and diffusion tensor images. The study consists of 60 ECT receiving inpatients suffering from major unipolar or bipolar depression. If successful, this line of research should lead to a better tolerability of ECT by aiding in the complex decision making process involved in prescribing ECT as well as the parameter setting within a treatment course (e.g. uni- vs. bilateral).