Capsule networks for segmentation of small intravascular ultrasound image datasets

1

Intravascular ultrasound (IVUS) is crucial for assessing vessel morphologies and improving treatment planning for coronary interventions.
2

Manual segmentation of IVUS images is time-consuming and relies heavily on physician experience, highlighting the need for automated methods.
3

Capsule networks, a novel architecture, group neurons into capsules to better capture part-whole relationships in image data.
4

This study presents an optimized capsule network for segmenting lumen and vessel wall in IVUS images, aiming to improve performance on small datasets.
5

Experimental results indicate that capsule networks may outperform traditional CNNs in segmenting IVUS images, particularly with limited training data.

1