Classification of neck tissues in OCT images by using convolutional neural network.

Area of Science:

• Medical Imaging
• Surgical Technology
• Artificial Intelligence in Medicine

Background:

• Accurate identification of surrounding neck tissues is critical during thyroid surgery.
• Optical coherence tomography (OCT) offers high-resolution, non-invasive imaging with potential for intraoperative tissue differentiation.
• Automated classification of OCT images can enhance surgical precision and safety.

Purpose of the Study:

• To develop and evaluate an automated system for classifying neck tissues using optical coherence tomography (OCT) images.
• To compare the performance of different convolutional neural network (CNN) architectures for this classification task.
• To investigate the effect of transfer learning on the accuracy and training efficiency of OCT image classification.

Main Methods:

• Collected OCT images of five distinct neck tissue types using a custom swept-source OCT (SS-OCT) system.
• Built a dataset for training and testing three CNN models: LeNet, VGGNet, and ResNet.
• Evaluated the impact of transfer learning on model performance, focusing on classification accuracy and training time.

Main Results:

• ResNet achieved the highest classification accuracy among the tested neural networks for neck tissue OCT images.
• Transfer learning did not significantly improve classification accuracy.
• Transfer learning demonstrated a tendency to accelerate network convergence and reduce overall training time.

Conclusions:

• Convolutional neural networks, particularly ResNet, show strong potential for automated neck tissue classification in OCT images.
• While transfer learning offers training efficiency benefits, its impact on classification accuracy for this specific OCT dataset was minimal.
• Automated OCT image analysis holds promise for improving intraoperative decision-making in thyroidectomy.