Download PDFOpen PDF in browser

Prediction of the Dosage of the Electric Stimulus Needed for Electroconvulsive Therapy (ECT) Based on Patient’s Pre-Ictal EEG Using Artificial Intelligence

EasyChair Preprint 9991

4 pagesDate: April 25, 2023

Abstract

One of the most effective and rapid treatments for MDD is Electroconvulsive Therapy (ECT). However, cognitive adverse effects remain a great risk among patients undergoing ECT. These side effects are robustly tied to the dosage of the electric stimulus given to the patient. Two methods are currently used to determine an accurate dosage: the age-based method and the titration method. Furthermore, electroencephalograms (EEG) are done during an ECT session, to assess the adequacy of the treatment. Therefore, Artificial Intelligence (AI) could offer a third way, by analyzing the EEG before the shock is administered (called the pre-ictal EEG), using deep learning algorithms, to determine the adequate dosage of the electric stimulus needed. Once the EEG signals were decomposed using Fast Fourier Transform (FFT), we fed them into the Fuzzy Causal Effect Variational Auto Encoder (FCEVAE) deep learning algorithm. We implemented an FCEVAE model to identify patterns in patients' pre-ictal EEGs that lead to positive or negative outcomes of the ECT session. These outcomes were determined by the clinician in charge of the ECT session, based on the EEG assessment. A total of 470 EEGs were collected. The FCEVAE seems able to predict individualized ECT dosages based on the patient’s pre-ictal EEG. The FCEVAE model had an overall accuracy of 90.33%, as measured by the root mean square measure. The use of FCEVAE seems promising in the field of EEG analysis and ECT, although further research is needed to optimize the model and its clinical applications.

Keyphrases: Artificial Intelligence, Causal Transformers, ECT, EEG, Electroconvulsive Therapy, Electroencephalogram, neural network, transformers

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
@booklet{EasyChair:9991,
  author    = {François-Xavier Roucaut and Usef Faghihi and Cyrus Kalantarpour},
  title     = {Prediction of the Dosage of the Electric Stimulus Needed for Electroconvulsive Therapy (ECT) Based on Patient’s Pre-Ictal EEG Using Artificial Intelligence},
  howpublished = {EasyChair Preprint 9991},
  year      = {EasyChair, 2023}}
Download PDFOpen PDF in browser