Automated irreversible electroporated region prediction using deep neural network, a preliminary study for treatment...

Area of Science:

• Biomedical Engineering
• Computational Biology
• Oncology

Background:

• Irreversible electroporation (IRE) is a cancer treatment aiming to maximize tumor cell death while sparing healthy tissue.
• Finite element analysis is commonly used for IRE simulations but has limitations.
• Accurate prediction of the irreversible electroporated region is crucial for effective treatment planning.

Purpose of the Study:

• To develop and evaluate a deep neural network (DNN) model for predicting irreversible electroporated regions in IRE cancer treatment.
• To assess the performance of the U-Net architecture in forecasting treatment zones for improved IRE planning.

Main Methods:

• Simulated IRE using COMSOL Multiphysics, incorporating electric conductivity changes.
• Generated datasets of electric field and cell kill probability distributions using eight pulses (100 μs width, 1 Hz frequency, varying voltages).
• Trained a U-Net architecture using a 90% cell kill probability contour as a mask for predicting irreversible electroporated regions.

Main Results:

• The U-Net model achieved an average DICE coefficient of 0.96 on test data.
• The average accuracy of the U-Net model in predicting irreversible electroporated regions was 0.97.
• Demonstrated high precision in identifying the extent of tissue affected by IRE.

Conclusions:

• This study presents the first known application of the U-Net architecture for predicting irreversible electroporated regions in IRE.
• The findings provide strong evidence for the efficacy of U-Net in IRE treatment planning.
• The developed DNN approach offers a promising tool for enhancing the accuracy and safety of irreversible electroporation cancer therapy.