LVADet: A Computer Vision-Based System for Detecting Lymphatic Vessels in Indocyanine Green Images

Yen-L. Tai,Kai-Y. Cheng,Yu-C. Tseng,Yi-T. Chen,Chih-S. Lai
DOI: https://doi.org/10.1109/access.2024.3510714
IF: 3.9
2024-12-14
IEEE Access
Abstract:Lymphaticovenous anastomosis (LVA) is a critical surgical approach for managing limb lymphedema, proving particularly effective in mild cases where lymphatic vessels are visible via indocyanine green (ICG) imaging. However, in advanced cases, substantial lymphatic obstructions manifest as stardust patterns in ICG images, posing diagnostic challenges. Traditionally, surgeons have relied on manually extrapolating the position and orientation of lymphatic vessels from the unaffected limb to the affected one based on the assumption of bodily symmetry. To streamline this process and enhance accuracy, we developed the LVA detection model (LVADet), which employs computer vision-based artificial intelligence techniques to automate the detection of lymphatic vessels. LVADet uses a segmentation algorithm to identify lymphatic vessels on ICG images of the unaffected limb and performs keypoint detection on both limbs. Through Procrustes analysis, keypoints from both limbs are used to construct an affine transformation matrix, enabling the projection of segmented vessels from the unaffected limb onto the image of the affected limb. We tested the model using a dataset of 52 lymphedema cases, with 42 used for training and 10 for testing. This included high-quality annotations of segmentations, keypoints, and incisional wounds. Our findings, supported by rigorous evaluation and ablation studies, validate the model's design choices. Furthermore, we integrated LVADet into a prototype system using Microsoft HoloLens, facilitating real-time identification of lymphatic vessels during LVA procedures. This integration significantly reduces labor and time in the preparatory phases of surgery and shows the potential to mark a substantial advancement in surgical treatments for lymphedema.
computer science, information systems,telecommunications,engineering, electrical & electronic
What problem does this paper attempt to address?