Optimal autoregressive orders for myopathic electromyograms.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuromuscular Disorders

Background:

• Electromyography (EMG) signals are crucial for diagnosing neuromuscular conditions.
• Accurate modeling of EMG signals is essential for effective analysis.
• Autoregressive (AR) models are commonly used for time-series signal representation.

Purpose of the Study:

• To identify the optimal autoregressive model order for varying-length electromyography (EMG) epochs in myopathic subjects.
• To evaluate the performance of different autoregressive orders in representing EMG signal complexity.
• To analyze the distribution and characteristics of optimal autoregressive orders.

Main Methods:

• Electromyography (EMG) signal epochs from myopathic subjects were analyzed.
• Autoregressive (AR) models with orders ranging from 1 to 100 were applied.
• The Minimum Description Length (MDL) criterion was used to select the optimal AR order for each epoch.
• Probability density functions (PDFs) were fitted to the distribution of optimal AR orders.

Main Results:

• The Minimum Description Length (MDL) criterion effectively determined the optimal autoregressive (AR) order for EMG epochs.
• The distribution of optimal AR orders was best described by a lognormal function.
• A linear relationship was observed between the mean of the optimal AR orders and the epoch length.

Conclusions:

• The optimal autoregressive order for myopathic EMG signals is dependent on epoch length.
• The lognormal distribution provides a robust model for the variability of optimal AR orders.
• These findings contribute to a more precise characterization and analysis of EMG signals in myopathy.