Gradient Segmentation for Enhanced Visualization of blood vessels in medical imaging

Deepak Menon K, Rashmitha R Mabiyan

Keywords: Arteriovenous fistula, Gradient Segmentation, Artificial Intelligence

Abstract: Traditional medical imaging segmentation classifies anatomical structures distinctly but fails to capture gradual transitions. We propose a gradient segmentation approach for smoother, more realistic boundaries, improving visualization and aiding clinical decisions. Using arteriovenous fistula (AVF) as an example, we modify neural network segmentation models to retain probability scores, applying gradient-based color mapping—red for arteries, blue for veins, and blended gradients for overlapping regions. This technique enhances vessel visualization and can be extended to other anatomical structures, providing a more intuitive approach to medical imaging.

Submission Number: 10