Improving visualisation of bronchi in three-dimensional rendering of CT data

The medical imaging system Sectra PACS from Sectra Imtec includes a 3 dimensional setting which you can use for visualising image stacks from e.g. computed tomography. Numerous structures of anatomy of human body could be visualised in the Three dimensional mode, but visualisations of the bronchial tree of the lungs hardly ever become good enough to be of help. The purpose of this report was to take a look at methods for bettering such visualisations. Different techniques had been researched, examined and tested. The fact that most effort was required for small structures with sizes just like the resolution of the images made things a little more intricate. A technique classifying neighbourhoods depending on local structure appeared as most promising, and was utilized as foundation for a suggested algorithm. It produces a mask which represents the existence of bronchi, permitting the concealing of uninteresting structures in its proximity. The algorithm was then applied in order that it could possibly be tested along with the present system. The process was found to be effective and was able to identify the smaller tubes of the bronchial tree and output the desired classification mask. Its effectiveness was a bit decreased by issues associated with speed…

Contents: Improving visualisation of bronchi in three-dimensional rendering of CT data

1 Introduction
1.1 Background
1.2 Goals
1.3 Outline of report
1.4 Overview of method
2 CT imaging of human lungs
2.1 The human lung
2.2 Computed tomography
2.2.1 CT
2.2.2 Reconstruction
2.2.3 Partial volume effects
2.2.4 Facts in CT data
2.2.5 Configuration choices
3 Visualisation of medical data
3.1 PACS and DICOM
3.1.1 PACS
3.1.2 DICOM
3.2 3 Volume rendering
3.2.1 Transfer functions
3.2.2 The three dimensional viewing mode
3.3 Dataset complications
4 Methods for image volume enhancement
4.1 Scale space
4.2 Density space
4.3 Decomposable convolution kernels
4.4 Simple filters
4.4.1 Edge sharpening filters
4.4.2 Median filtering
4.5 Tensors and similar constructs for local structure
4.5.1 Tensors
4.5.2 Simple signals and function rank
4.5.3 Adaptivefilters
4.5.4 Usingeigenvalues
4.5.5 Calculation of eigenvalues
4.5.6 Gradienttensors
4.5.7 Hessian matrix for local structure
4.5.8 Calculating a tubeness measure from eigenvalues
4.6 Medialness filters
4.7 Quadrature filters
4.8 Segmentation by region growing or morphological techniques
4.9 Diffusion based methods
5 Improving the visualisation
5.1 Datasets
5.2 Tools
5.3 Evaluation
5.4 Systemintegration
6 Evaluation results
6.1 Filteringmethods
6.1.1 Simplefilters
6.1.2 Gradienttensors
6.1.3 Hessianmethod
6.1.4 Combining gradient tensors and Hessian matrices
6.1.5 Medialness filters
6.1.6 Quadrature filters
6.1.7 Region growing and diffusion
6.2 Variationsoffilteringmethods
6.2.1 Handlinganisotropicdata
6.2.2 Processing data at lower resolutions
6.2.3 Densityspace
6.2.4 Blockprocessing
6.3 Applicationtodata
6.4 Conclusions of evaluation
7 Implementation
7.1 General implementation considerations
7.2 Toolkitsandlibraries
7.3 ImplementationwithITK
7.3.1 ThepipelinestructureofITK
7.3.2 Effects of block splitting
7.3.3 Scalespace
7.3.4 Ellipsoids
7.4 Integration with Sectra PACS
7.5 Speeding up eigenvalue calculation
7.6 The resulting algorithm
7.7 Processing times
7.8 Tests on more data sets
7.8.1 Applicability to datasets
7.8.2 Handling large bronchi
7.8.3 Tubemodel
8 Discussion, conclusion and summary
8.1 Discussion
8.1.1 Performance….

Source: Linköping University

Download URL 2: Visit Now

Leave a Comment