Predicting Prognostic Effects of Acupuncture for Depression Using the Electroencephalogram

Area of Science:

• Neuroscience and Computational Psychiatry
• Integrative Medicine and Health Informatics

Background:

• Depression poses a significant public health challenge.
• Acupuncture is a complementary therapy for depression, but predicting its effectiveness is crucial for timely interventions.
• Electroencephalogram (EEG) is a valuable tool for assessing acupuncture's therapeutic impact.

Purpose of the Study:

• To develop a novel framework for predicting the prognostic effects of acupuncture in depression patients using EEG data.
• To identify key EEG features that indicate treatment response.
• To build and evaluate machine learning models for predicting acupuncture efficacy.

Main Methods:

• Feature selection using Max-Relevance and Min-Redundancy (mRMR) on EEG lead-rhythm features.
• Calculating the reduction rate of Hamilton Depression Rating Scale (HAMD) scores as a measure of prognostic effect.
• Developing predictive models using five machine learning algorithms, including Support Vector Machine with Gaussian Kernel (SVM-RBF).

Main Results:

• Nonlinear machine learning models outperformed linear models in predicting acupuncture's prognostic effects.
• The SVM-RBF model achieved the best and most stable performance, with accuracy of 84.61% and F1 score of 86.67%.
• mRMR feature selection effectively identified relevant EEG lead-rhythm features, contributing to model accuracy.

Conclusions:

• The proposed framework effectively predicts the prognostic effects of acupuncture for depression using EEG.
• This approach can be integrated into intelligent medical systems to aid physicians in treatment decisions.
• Informed prognostic predictions can lead to better interventions and potentially reduce adverse events in mental health patients.