Ultrasound volume projection image quality selection by ranking from convolutional RankNet

Highlights•Considers the selection of the ultrasound spine images with large number of similarities as a ranking problem. •Selects the best image by ranking them in sequence. •Designs a specific CNN architecture for the feature extraction to replace the artificial neural network in the conventional...

Full description

Saved in:
Bibliographic Details
Published in:Computerized medical imaging and graphics 2021-04, Vol.89, p.101847-101847, Article 101847
Main Authors: Lyu, Juan, Ling, Sai Ho, Banerjee, S, Zheng, J.Y, Lai, K.L, Yang, D, Zheng, Y.P, Bi, Xiaojun, Su, Steven, Chamoli, Uphar
Format: Article
Language:English
Subjects:
3D ultrasound imaging
Algorithms
Artificial neural networks
Backbone
Convolutional RankNet
Feature extraction
Image quality
Image selection
Imaging techniques
Inspection
Internal Medicine
Machine learning
Neural networks
Other
Radiation
Ranking
Scoliosis
Spine
Ultrasonic imaging
Ultrasound
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Highlights•Considers the selection of the ultrasound spine images with large number of similarities as a ranking problem. •Selects the best image by ranking them in sequence. •Designs a specific CNN architecture for the feature extraction to replace the artificial neural network in the conventional RankNet. •Adopts the hinge loss function rather than the cross-entropy loss function in the conventional RankNet, which has more discriminating ability when the estimated output scores of two images are very close.
ISSN:0895-6111
1879-0771
DOI:10.1016/j.compmedimag.2020.101847