Figure - PMC

Skip to main content

An official website of the United States government

Here's how you know

Here's how you know

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

View full-text article in PMC

. 2021 Jul 12;11:14306. doi: 10.1038/s41598-021-93202-y

Search in PMC
Search in PubMed
View in NLM Catalog
Add to search

© The Author(s) 2021

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

PMC Copyright notice

Qualitative nerve segmentation results. Qualitative nerve segmentation results under different image conditions (picture distance and brightness), with summary statistics and example images from all 5 cross-validation test folds. For each of the six image conditions labeled at the top of the figure, we show three results: a plot at the top with the mean DSC scores and 95% confidence intervals (as shown in Table 1), and in the two bottom boxes, corresponding example good (green) and worse (orange) results from each image condition to illustrate the range of model performance. For each vertical image pair in the two bottom boxes, the original image is on top and the overlaid predicted and ground-truth segmentations annotated by surgeons is on the bottom. In the overlaid segmentations, the intersection of the predicted and ground-truth segmentations (true positive) is green, false positive pixels are in yellow, and false negative pixels are in blue. Note that both far-away and close-up images and corresponding segmentations are shown post-crop with the cropping model for better segmentation visualization. In the good segmentation cases, the predicted segmentation masks track closely with the ground-truth and demonstrate successful learning of expected nerve morphology. Furthermore, notice that the algorithm performs well even in difficult scenarios such as very thin nerves (e.g. far-away, bright lighting), small nerves (far-away, medium lighting), and boundary details (close-up, dim lighting; close-up, medium lighting). The worse segmentation cases highlight challenging scenarios for the model, such as missing very small nerves (far-away, bright lighting), anatomies in better-lit portions distracting from a poorly-lit nerve (far-away, dim lighting), artifacts due to neighboring tissue texture (far-away, medium lighting), difficulty with irregular nerve shapes that deviate from more common, smooth nerve morphologies in the dataset (close-up, dim lighting; close-up, medium lighting), and light reflectance on the nerve leading to over-segmentation (close-up, bright lighting). The worse cases occur more frequently and severely in poorer distance and lighting conditions. These cases illustrate that capturing higher-quality images with respect to both picture distance and brightness is important for achieving better segmentation performance. Best viewed in color.