Automatic Assessment of Scoliosis Using 3D Ultrasound Imaging and Convolutional Neural Network

Banerjee, Sunetra

Automatic Assessment of Scoliosis Using 3D Ultrasound Imaging and Convolutional Neural Network

Banerjee, Sunetra

Permalink

Publication Type:: Thesis
Issue Date:: 2022

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Download contents and abstractAdobe PDF (779.91 kB)

Download thesisAdobe PDF (6.07 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Banerjee, Sunetra
dc.date.accessioned	2022-10-07T02:11:53Z
dc.date.available	2022-10-07T02:11:53Z
dc.date.issued	2022
dc.identifier.uri	http://hdl.handle.net/10453/162380
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	Scoliosis is a gradual 3D spinal deformation where the spine takes a lateral curvature, generating an angle in the coronal plane. This condition may start from adolescence and its diagnosis requires periodic detection. The conventional detection method involves measuring the Cobb angle in spine images obtained by anterior X-Ray scanning and frequent exposure to radiating imaging pose a radiation threat to the young patients. Ultrasound imaging of the spine is found to be safer than traditional radiographic modalities. For posterior ultrasound scanning, an alternate index called Ultrasound Curve Angle (UCA) was developed. The current practice of UCA angle measurement is manual and this research attempts to automate the process. The unique challenges in the research, i.e. automatic prognosis of scoliosis, are (a) dealing with spine images which have very high variability in shape, size, and location of bony features, and (b) handling images that are taken using Ultrasound which is inherent of low-contrast and plagued with speckle noises. The overall sequence of this research work is: {1} Manual selection of ultrasound images with best lateral bony features by experts, {2} Automatic segmentation of lateral bony features. Two novel deep learning segmentation techniques were iteratively developed during this research: (a) Light-Convolution Dense Selection (LDS) U-Net (b) Multi-Scale feature fusion Skip-Inception U-Net (SIU-Net), {3} Conversion of binary segmented images to RGB colour coded images and using them, automatically calculate the curvature angle using novel Centroid Pairing and Inscribed rectangle Slope (CPI-SLO) method, and {4} Validation with traditional manual methods. The dataset employed is the scanned images of 109 patients with different severity of scoliosis. The performance evaluation shows that this novel proposed method has a very good agreeability with the manual UCA method. The advantage of the research is that it reduces human intervention, making the scoliosis assessment process faster, more scalable, and affordable to economically challenged sections of society.	en_US.UTF-8
dc.format	Thesis (PhD)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/162380/5/02whole.pdf
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	au.edu.uts.lib/ppc
dc.rights	info:eu-repo/semantics/openAccess
dc.title	Automatic Assessment of Scoliosis Using 3D Ultrasound Imaging and Convolutional Neural Network	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

Scoliosis is a gradual 3D spinal deformation where the spine takes a lateral curvature, generating an angle in the coronal plane. This condition may start from adolescence and its diagnosis requires periodic detection. The conventional detection method involves measuring the Cobb angle in spine images obtained by anterior X-Ray scanning and frequent exposure to radiating imaging pose a radiation threat to the young patients. Ultrasound imaging of the spine is found to be safer than traditional radiographic modalities. For posterior ultrasound scanning, an alternate index called Ultrasound Curve Angle (UCA) was developed. The current practice of UCA angle measurement is manual and this research attempts to automate the process. The unique challenges in the research, i.e. automatic prognosis of scoliosis, are (a) dealing with spine images which have very high variability in shape, size, and location of bony features, and (b) handling images that are taken using Ultrasound which is inherent of low-contrast and plagued with speckle noises. The overall sequence of this research work is: {1} Manual selection of ultrasound images with best lateral bony features by experts, {2} Automatic segmentation of lateral bony features. Two novel deep learning segmentation techniques were iteratively developed during this research: (a) Light-Convolution Dense Selection (LDS) U-Net (b) Multi-Scale feature fusion Skip-Inception U-Net (SIU-Net), {3} Conversion of binary segmented images to RGB colour coded images and using them, automatically calculate the curvature angle using novel Centroid Pairing and Inscribed rectangle Slope (CPI-SLO) method, and {4} Validation with traditional manual methods. The dataset employed is the scanned images of 109 patients with different severity of scoliosis. The performance evaluation shows that this novel proposed method has a very good agreeability with the manual UCA method. The advantage of the research is that it reduces human intervention, making the scoliosis assessment process faster, more scalable, and affordable to economically challenged sections of society.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/162380