Tetrahedra), with triangle size independent of
Stradview contours, and the capability to interpolate
between very sparse contour sets.
Visualisation and analysis of 3D imaging data. • Automated extraction to multiple objects using
http://mi.eng.cam.ac.uk/Main/StradView data thresholds, but with tools to allow for sub-
threshold thin features and gaps in 3D, and
concavity detection and removal.
• Sub-pixel and sub-voxel accuracy throughout
this process, in contrast to surfaces based on
classifying voxels.
• Import objects from NIfTI and DICOM files and
surfaces from PLY files, and export intersections
and surfaces to PLY, NIfTI and NRRD label data.
• Also includes features for segmenting surfaces
using prior shape models.
• Drawing directly on surfaces to extract patches /
cut parts out for further analysis.
Concept
• Practical: makes very efficient use of memory
and graphics power and is highly optimised. It
can hence process large data sets (e.g. microCT)
as well as smaller ones (e.g. medical) quickly
even on fairly standard computers.
• Accurate: sub-pixel and sub-voxel resolution is
preserved throughout all of the analysis and
visualisation.
• Targeted: for visualising and analysing individual
3D (cross-sectional) data sets. Doesn’t do
everything: but does what it does well.
• Fine control: the interface is carefully designed,
but all useful options are exposed to the user. Visualise surfaces
• Unique: all features are designed and written in- • Includes advanced features such as ambient
house, not using third-party libraries. occlusion, soft and coloured translucent
shadowing, and material anisotropy.
• A 'clear' visualisation which clearly highlights
internal features within complicated surfaces.
• Correct translucency even for complicated,
multiple overlapping surfaces.
• Dynamic intersections with all data reslices.
Volume rendering
• High quality realistic visualisations, better than
normally available on low end hardware,
including radiant and ambient shadowing and
material properties.
• Very accurate, using real time cubic interpolation
Create surfaces from 3D data of data.
• Surfaces always derived via contours in the • Fast and efficient, can run on a fairly minimal
original data which can be manually edited even graphics card, reducing resolution automatically
if they were created automatically. if the memory is limited.
• Automated level-of-detail to ensure the • Volume rendering can be cropped to completely
minimum number of contour points for editing arbitrary surfaces, with sub-voxel accuracy.
whilst maintaining accuracy. Deletion and • A 'clear' visualisation (as for surfaces) shows
selection of contours in 2D and 3D. internal features very well.
• Surfaces created from contours contain very • Mixed volume rendering / surface visualisations.
regular triangles (Regularised Marching
Graham Treece, gmt11@cam.ac.uk, 1 December 2023 Page 1
Data handling
• Design for minimal memory use allows loading
and realistic processing of even large (>8Gb)
data sets.
• Can read many different data formats, including
DICOM, Mindways QCT, ITK, Analyze, NIfTI,
NRRD and many 2D image formats.
• Can be used to take cross-sectional image
sequences without any 3D locations and create
3D data from these, including when it is not on a
regular grid, e.g. irregular spacing or fan-scans.
• Can export irregular 3D data to regular 3D
Analyse data over surfaces resliced data, and regular data to NIfTI and
• Including Cortical Bone Mapping, the only NRRD formats.
software which implements this technique. • Can export data to sequences of cross-sectional
• Advanced features for extracting very thin (sub- 2D images and visualisations to images.
resolution) laminar surfaces from data. • Many functions can be scripted, allowing
• Can be used to create accurate geometry from automated processing and the creation of still
thin surfaces. image sequences for movies.
• Can be used to map thin surface features (e.g.
thickness / density) to the surface for
visualisation or later statistical analysis.
• Can also directly map imaging data (min / max /
average over a range) to the surface, which
allows segmentation on surfaces.
Reslice imaging data
• Can generate accurate cross-sectional images
(reslices) in any orientation.
• Thick slab reslices are also possible for e.g.
maximum intensity projections.
• All reslices can be cropped to completely
arbitrary surfaces.
Measurements
• Landmark placement on imaging data, reslices,
surfaces or even volume visualisations.
• Allows measurement of location, distance and
angles, and noting reference points.
• Export of this and summary surface information
(volume, cross-sectional areas, centres, etc) to
external files.
Graham Treece, gmt11@cam.ac.uk, 1 December 2023 Page 2
