Advanced

Segmentation, Analysis, and Modelling of Microstructure in Cortical Bone, based on X-ray Microtomography

Törnquist, Elin LU (2017) BMEM01 20171
Department of Biomedical Engineering
Abstract
The microstructure in cortical bone greatly affect the toughness of the bone and the crack propagation during fracture. The two main structural features, Haversian canals and osteons, have been previously studied in order to increase the knowledge of the biomechanics of bone. In studies where X-ray microtomography has been the imaging method of choice, the microstructure has been analysed based on manual segmentation, a method which is difficult and tedious on larger sample volumes.

X-ray microtomograms, or $\mu$CT images, are gray scale 2D images, which can be rendered into a 3D volume, where the pixel intensity corresponds to the absorption coefficient of the material in the object. Material that absorbs a lot of X-ray radiation will... (More)
The microstructure in cortical bone greatly affect the toughness of the bone and the crack propagation during fracture. The two main structural features, Haversian canals and osteons, have been previously studied in order to increase the knowledge of the biomechanics of bone. In studies where X-ray microtomography has been the imaging method of choice, the microstructure has been analysed based on manual segmentation, a method which is difficult and tedious on larger sample volumes.

X-ray microtomograms, or $\mu$CT images, are gray scale 2D images, which can be rendered into a 3D volume, where the pixel intensity corresponds to the absorption coefficient of the material in the object. Material that absorbs a lot of X-ray radiation will show up as high intensity pixel whilst material that absorbs little will show up as low intensity pixel, just as in a normal radiogram. $\mu$CT is best used on samples containing structures with varying absorbtion properties as this will enable good contrast between the structures.

In this project, a semi-automatic method for segmenting the microstructures Haversian canals and osteons in $\mu$CT images of cortical bovine bone was implemented. Based on this segmentation, a simplified model was built for future Finite Element Modelling. K-means clustering with nine clusters was chosen as segmentation method. By identifying which clusters correspond to what pixel intensities, and hence to what tissue type, it was possible to segment out Haversian canals, and osteons in the $\mu$CT images. The simplified model was based on principal component analysis of the segmented canals in 3D, and circle fitting on Haversian canals, and osteons in 2D.

The generated segmentations provided enhanced visibility of the microstructure. Based on porosity, and volume analysis, the segmentation pipeline gave good results for the Haversian canals, but was less accurate in differentiating between osteonal tissue and tissue with similar absorption properties embedded in the interstitial bone. Porosity measurements on the segmentations corresponded with previous studies. The radii of the osteons and the Haversian canals, generated by the simplified model, agreed well with previous studies, and the method overcame the issue with unclear osteonal boundaries by using the segmentation of the Haversian canals as a base. (Less)
Popular Abstract (Swedish)
Vårt skelett är uppbyggt av olika typer av benvävnad, vilka alla har komplex struktur och olika egenskaper. Benvävnadens struktur motverkar på olika sätt att en spricka växer till en fraktur. Exakt hur detta går till är ännu inte förstått men man vet att strukturerna, från på nanometernivå till i full skala, på olika sätt bidrar till att stå emot benbrott.
I det här projektet utvecklas en ny metod för att titta på storleken och riktningen hos den kompakta benvävnadens mikrostruktur, alltså de benstrukturer som är mindre än ett hundratal mikrometer. Metoden gör att analysen går snabbare och resultatet blir lättare att jämföra, vilket förenklar för forskare som vill förstå hur benbrott sker.
Please use this url to cite or link to this publication:
author
Törnquist, Elin LU
supervisor
organization
course
BMEM01 20171
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Cortical bone, Biomechanics, Segmentation, K-mean clustering, Microstructure, X-ray Microtomography
language
English
additional info
2017-11
id
8912841
date added to LUP
2017-06-27 09:53:10
date last changed
2017-06-27 09:53:10
@misc{8912841,
  abstract     = {The microstructure in cortical bone greatly affect the toughness of the bone and the crack propagation during fracture. The two main structural features, Haversian canals and osteons, have been previously studied in order to increase the knowledge of the biomechanics of bone. In studies where X-ray microtomography has been the imaging method of choice, the microstructure has been analysed based on manual segmentation, a method which is difficult and tedious on larger sample volumes. 

X-ray microtomograms, or $\mu$CT images, are gray scale 2D images, which can be rendered into a 3D volume, where the pixel intensity corresponds to the absorption coefficient of the material in the object. Material that absorbs a lot of X-ray radiation will show up as high intensity pixel whilst material that absorbs little will show up as low intensity pixel, just as in a normal radiogram. $\mu$CT is best used on samples containing structures with varying absorbtion properties as this will enable good contrast between the structures.

In this project, a semi-automatic method for segmenting the microstructures Haversian canals and osteons in $\mu$CT images of cortical bovine bone was implemented. Based on this segmentation, a simplified model was built for future Finite Element Modelling. K-means clustering with nine clusters was chosen as segmentation method. By identifying which clusters correspond to what pixel intensities, and hence to what tissue type, it was possible to segment out Haversian canals, and osteons in the $\mu$CT images. The simplified model was based on principal component analysis of the segmented canals in 3D, and circle fitting on Haversian canals, and osteons in 2D.

The generated segmentations provided enhanced visibility of the microstructure. Based on porosity, and volume analysis, the segmentation pipeline gave good results for the Haversian canals, but was less accurate in differentiating between osteonal tissue and tissue with similar absorption properties embedded in the interstitial bone. Porosity measurements on the segmentations corresponded with previous studies. The radii of the osteons and the Haversian canals, generated by the simplified model, agreed well with previous studies, and the method overcame the issue with unclear osteonal boundaries by using the segmentation of the Haversian canals as a base.},
  author       = {Törnquist, Elin},
  keyword      = {Cortical bone,Biomechanics,Segmentation,K-mean clustering,Microstructure,X-ray Microtomography},
  language     = {eng},
  note         = {Student Paper},
  title        = {Segmentation, Analysis, and Modelling of Microstructure in Cortical Bone, based on X-ray Microtomography},
  year         = {2017},
}