Using Deep Learning to Detect and Evaluate Blood Flow of Parathyroid Glands During Endoscopic Thyroid Surgery

What this Trial Is About

Researchers are developing a deep convolutional neural network DCNN to help detect the parathyroid gland PTG and assess its blood supply during endoscopic thyroidectomy. The study addresses the challenge that the PTGs location and appearance vary, which can make it difficult to identify and preserve during surgery. Mistaken removal or blood supply disruption of the PTG can cause hypoparathyroidism, requiring calcium or vitamin D supplements. The goal is to improve surgical outcomes by supporting surgeons with advanced imaging analysis. The study involves training the DCNN using a large number of endoscopic images to build a model that assists surgeons in identifying the PTG during surgery. Additionally, the DCNN will be trained to predict blood perfusion of the PTG by comparing its assessments to indocyanine green fluorescence angiography, which is currently used but has risks like allergic reactions and requires repeated injections. This research aims to create endoscopic tools for better PTG detection and perfusion evaluation. Participants are patients undergoing endoscopic thyroidectomy. Researchers will collect and analyze endoscopic images to train and test the DCNN. The primary outcome measure is the area under the receiver operating characteristic curve evaluated over 3 years, which reflects the DCNNs accuracy. This study is observational, with no experimental treatments, focusing on developing diagnostic technology to reduce postoperative hypoparathyroidism risk.

DCNN Developed for Detection and Assessing the Perfusion of PTG