Multi-View Consistent Wound Segmentation With Neural Fields

TL;DR

This paper introduces WoundNeRF, a neural field-based method for consistent multi-view 3D wound segmentation from 2D images, demonstrating improved accuracy over existing approaches.

Contribution

The paper presents WoundNeRF, a novel NeRF SDF-based approach for multi-view consistent wound segmentation, addressing the challenge of inferring 3D structures from 2D images.

Findings

WoundNeRF outperforms state-of-the-art Vision Transformer networks.

WoundNeRF provides more accurate and consistent 3D wound segmentations.

The method demonstrates robustness with automatically generated annotations.

Abstract

Wound care is often challenged by the economic and logistical burdens that consistently afflict patients and hospitals worldwide. In recent decades, healthcare professionals have sought support from computer vision and machine learning algorithms. In particular, wound segmentation has gained interest due to its ability to provide professionals with fast, automatic tissue assessment from standard RGB images. Some approaches have extended segmentation to 3D, enabling more complete and precise healing progress tracking. However, inferring multi-view consistent 3D structures from 2D images remains a challenge. In this paper, we evaluate WoundNeRF, a NeRF SDF-based method for estimating robust wound segmentations from automatically generated annotations. We demonstrate the potential of this paradigm in recovering accurate segmentations by comparing it against state-of-the-art Vision Transformer networks and conventional rasterisation-based algorithms. The code will be released to facilitate further development in this promising paradigm.