Core Needle Biopsy Guidance Based on Tissue Morphology Assessment with AI-OCT Imaging

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Artificial Intelligence

Background:

• Up to 40% of biopsy cores lack diagnostic value due to adipose or necrotic tissue.
• Accurate real-time tissue analysis before biopsy is clinically significant.

Purpose of the Study:

• To develop and evaluate a combined optical imaging/artificial intelligence (OI/AI) technique for real-time tissue morphology analysis at the biopsy needle tip.
• To improve the selection of biopsy locations and enhance diagnostic yield.

Main Methods:

• Utilized micron-scale-resolution optical coherence tomography (OCT) imaging via a minimally invasive needle probe.
• Employed a convolutional neural network (CNN)-based artificial intelligence (AI) software for automated analysis of OCT images.
• Trained the AI model using annotated OCT images as ground truth in a rabbit cancer model.

Main Results:

• The AI model achieved performance comparable to human experts in tissue classification.
• Excellent tissue segmentation accuracy of approximately 99% was obtained.
• Achieved over 84% correlation accuracy for tumor and non-tumor classification.

Conclusions:

• The combined OI/AI technique provides real-time analysis of tissue morphology, enabling more adequate biopsy site selection.
• This technology demonstrates high accuracy in tissue segmentation and tumor classification, with potential to reduce inadequate biopsy samples.