Tetanus Severity Classification in Low-Middle Income Countries through ECG Wearable Sensors and a 1D-Vision...

Area of Science:

• Medical Informatics
• Machine Learning
• Cardiology

Background:

• Tetanus is a severe bacterial infection impacting the nervous system, particularly prevalent in low- and middle-income countries.
• Autonomic nervous system (ANS) dysfunction is common in severe tetanus, necessitating continuous vital sign monitoring.
• Wearable electrocardiogram (ECG) sensors offer a practical alternative to traditional bedside monitors for vital sign assessment.

Purpose of the Study:

• To develop a machine learning approach for tetanus severity classification using ECG data without time series imaging.
• To evaluate the performance of a 1D-Vision Transformer model for this classification task.
• To achieve comparable or improved classification performance against existing methods.

Main Methods:

• Utilized a novel 1D-Vision Transformer model to analyze 1D ECG signals.
• Extracted global information directly from ECG data, avoiding image-based processing.
• Compared the proposed model against 1D-CNN, 2D-CNN, and 2D-CNN + Dual Attention models.

Main Results:

• The 1D-Vision Transformer model achieved an F1 score of 0.77 ± 0.06.
• The model demonstrated strong performance with precision of 0.70 ± 0.09, recall of 0.89 ± 0.13, and accuracy of 0.82 ± 0.06.
• Achieved an Area Under the Curve (AUC) of 0.84 ± 0.05, outperforming other evaluated methods.

Conclusions:

• The 1D-Vision Transformer presents a pioneering and effective method for classifying tetanus severity directly from 1D ECG signals.
• This approach offers a more efficient alternative to time-consuming ECG time series imaging techniques.
• The model shows significant potential for improving the diagnosis and management of tetanus, particularly in resource-limited settings.