Related Experiment Video
Updated: Sep 7, 2025
05:44
Performing In Vivo and Ex Vivo Electrical Impedance Myography in Rodents
Published on: June 8, 2022
3.1K
Using machine learning algorithms to enhance the diagnostic performance of electrical impedance myography
Sarbesh R Pandeya1, Janice A Nagy1, Daniela Riveros1
1Department of Neurology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts, USA.
Muscle & Nerve
|June 21, 2022
Summary
Machine learning (ML) with multifrequency electrical impedance myography (EIM) significantly improves disease classification in mice compared to single-value EIM analysis. This enhanced diagnostic capability holds promise for future clinical applications.
Area of Science:
- Biomedical Engineering
- Machine Learning in Diagnostics
- Musculoskeletal Research
Background:
- Electrical impedance myography (EIM) is a non-invasive technique to assess muscle health.
- Current EIM analysis often relies on single-frequency values, potentially limiting diagnostic accuracy.
- Advancements in machine learning (ML) offer opportunities to enhance complex data analysis in biomedical applications.
Purpose of the Study:
- To evaluate the diagnostic performance of ML algorithms using multifrequency EIM data.
- To compare the classification accuracy of ML-based multifrequency EIM against single-frequency EIM analysis.
- To determine the potential of ML-enhanced EIM for improved disease diagnosis in muscle pathologies.
Main Methods:
- EIM data (resistance, reactance, phase) were collected across multiple frequencies from excised gastrocnemius muscles of diseased and wild-type mice.
- A random forest ML algorithm was trained and tested using the comprehensive multifrequency EIM dataset.
- Classification performance was assessed by comparing ML model outputs against a single 50 kHz EIM phase value using receiver-operating characteristic curves and area under the curve (AUC).
Main Results:
- The ML model incorporating multifrequency EIM data demonstrated superior classification performance compared to the single 50 kHz EIM phase value.
- For classifying all diseases versus wild-type, the ML model achieved an AUC of 0.94, significantly outperforming the 50 kHz phase AUC of 0.52.
- In discriminating Amyotrophic Lateral Sclerosis (ALS) models from wild-type, the ML model yielded an AUC of 0.99, compared to 0.79 for the 50 kHz phase.
Conclusions:
- Multifrequency EIM combined with ML significantly enhances classification accuracy for muscle diseases over single-frequency EIM.
- ML-driven analysis of multifrequency EIM data represents a powerful tool for improving diagnostic outcomes.
- These findings support the integration of ML-based multifrequency EIM in future diagnostic strategies for both research and clinical settings.
