Predicting the Risk of Depression Based on ECG Using RNN

Area of Science:

• Cardiology
• Psychiatry
• Artificial Intelligence

Background:

• Depression is linked to distinct heartbeat patterns, including Premature Ventricular Contractions (PVCs).
• Electrocardiogram (ECG) data offers a potential non-invasive method for assessing cardiac anomalies associated with mental health conditions.

Purpose of the Study:

• To develop and validate a deep learning model for predicting depression risk using ECG signals.
• To differentiate between normal, abnormal, and potentially risky heartbeats, specifically PVCs.

Main Methods:

• Utilized a Recurrent Neural Network (RNN) with a Long Short-Term Memory (LSTM) autoencoder architecture.
• Trained and tested the model on a dataset of 5000 ECG samples, classifying heartbeats as normal, abnormal, or PVC.

Main Results:

• Achieved 97.24% accuracy in predicting normal heartbeats and 100% accuracy for PVC heartbeats.
• Demonstrated effective differentiation between normal, abnormal, and risky heartbeats through low training loss values (best losses: 5.71 for normal, 33.36 for abnormal, 34.78 for risky).

Conclusions:

• The developed RNN-LSTM model shows high accuracy in identifying cardiac irregularities potentially indicative of depression.
• Further research with larger datasets may enhance the model's predictive capabilities for depression risk assessment using ECG.