Machine learning for intraoperative prediction of viability in ischemic small intestine

Area of Science:

• Biomedical Engineering
• Surgical Innovation
• Machine Learning in Medicine

Background:

• Accurate assessment of intestinal viability is critical for surgical decisions in acute intestinal ischemia (AMI).
• Current diagnostic methods, including initial laparotomy (50% accuracy), have limitations, with second-look laparotomy as the gold standard (87-89% accuracy).
• High mortality rates (30-93%) persist for AMI patients, highlighting the need for improved diagnostic tools.

Purpose of the Study:

• To investigate the efficacy of machine learning algorithms for classifying intestinal viability and histological grading.
• To compare the accuracy of feedforward neural networks (FNN) using single bioimpedance measurements versus recurrent neural networks with long short-term memory units (LSTM-RNN) using historical bioimpedance data.
• To evaluate multivariate time-series bioimpedance sensor data for assessing intestinal viability in a porcine jejunum model.

Main Methods:

• Acquisition of multivariate time-series bioimpedance sensor data from pig jejunum under various conditions (perfused, ischemic, reperfused).
• Comparison of classification accuracy between FNN models utilizing end-point bioimpedance measurements and LSTM-RNN models incorporating data history.
• Analysis of bioimpedance parameters over extended periods of ischemia (1-16 hours) and reperfusion (1-8 hours).

Main Results:

• Feedforward neural networks achieved accuracies comparable to clinical findings using single bioimpedance measurements.
• Recurrent neural networks with long short-term memory units demonstrated higher accuracies when analyzing a sequence of bioimpedance data history.
• Machine learning models effectively classified intestinal viability and histological grading based on bioimpedance sensor data.

Conclusions:

• Intraoperative bioimpedance measurements combined with machine learning significantly enhance the accuracy of intestinal viability assessment.
• The improved accuracy of intraoperative assessment holds the potential to reduce the high mortality and morbidity rates associated with acute intestinal ischemia.
• LSTM-RNN models analyzing bioimpedance data history show promise for real-time, accurate intraoperative decision-making in intestinal ischemia cases.