Deep Learning-Based AI Model for Brain Tumor Segmentation in Digital Pathology and Terahertz Imaging

Abstract

This study presents a deep learning-based AI model for brain tumor segmentation in digital pathology images. Using a transgenic mouse model and H&E-stained images, we developed and trained the model with DEEP:PHI, employing U-Net and attention U-Net architectures. The AI model facilitates accurate cancer detection, contributing to terahertz imaging-based diagnostics and enhancing real-time surgical decision-making with minimal pathologist intervention. © 2025 The Author(s)