Small intestinal viability assessment using dielectric relaxation spectroscopy and deep learning.

Area of Science:

• Biomedical Engineering
• Surgical Innovation
• Gastroenterology

Background:

• Intestinal ischemia poses critical surgical challenges in determining resection margins.
• Distinguishing between viable and non-viable bowel tissue within 3-4 hours of ischemia is difficult with current methods.
• Accurate intraoperative assessment of intestinal viability is crucial for patient outcomes.

Purpose of the Study:

• To investigate the utility of dielectric relaxation spectroscopy for assessing intestinal viability.
• To develop a machine learning-based method for classifying intestinal tissue under various ischemia/reperfusion conditions.
• To provide surgeons with a tool for rapid and accurate intraoperative decision-making regarding intestinal resection.

Main Methods:

• Collected permittivity data from porcine intestinal segments subjected to varying degrees of ischemia and reperfusion.
• Analyzed dielectric constant and conductivity changes in relation to tissue condition.
• Developed and applied machine learning models to classify tissue viability based on frequency-dependent dielectric properties.

Main Results:

• Dielectric parameters (dielectric constant and conductivity) clearly differentiated between healthy, ischemic, and reperfused intestinal segments.
• Machine learning models achieved high classification accuracy: 98.7% during ischemia and 96.2% during reperfusion.
• Permittivity measurements combined with machine learning offer a sensitive method for evaluating intestinal viability.

Conclusions:

• Dielectric relaxation spectroscopy is a promising technique for assessing intestinal viability.
• Machine learning models trained on dielectric data can accurately guide intraoperative surgical decisions.
• This approach offers a fast and reliable method for evaluating questionable intestinal segments during surgery.