Electric impedance spectroscopy feature extraction for tissue classification with electrode embedded surgical needles
Computers in Biology and Medicine
|June 21, 2021
Summary
This study introduces a new method for electric impedance sensing surgical tools to identify tissues. The proposed technique effectively reduces features for accurate tissue classification during surgery.
Area of Science:
- Biomedical Engineering
- Surgical Technology
- Medical Devices
Background:
- Electric impedance sensing shows promise for surgical tissue identification.
- Developing effective feature extraction and classification methods is crucial for these tools.
- Distinguishing between various tissue types is essential for surgical accuracy.
Purpose of the Study:
- To explore feature extraction and classification techniques for electric impedance data.
- To propose a modified forward stepwise method for feature selection.
- To evaluate the effectiveness of these methods on diverse biological tissue samples.
Main Methods:
- Applied various feature extraction techniques and classification methods to electric impedance data.
- Developed a novel scoring metric for feature selection based on coefficient of variation and overlapping index.
- Tested methods on spectral data from 132 samples across 6 tissue types (bovine, poultry, canine).
Main Results:
- Successfully identified distinct electric impedance spectra features for biological tissues.
- Boruta feature extraction with Random Forest classifier yielded high accuracy but minimal feature reduction.
- The proposed method reduced features to an average of 5.8 per classifier.
Conclusions:
- Electric impedance sensing offers potential for surgical tissue identification.
- The proposed feature selection method effectively reduces model complexity.
- These techniques could aid in minimally invasive cancer surgery by improving lesion targeting.
