The Multi-Spectral Imaging via Computed Tomography (MUSIC) dataset in a two-part (2D- and 3D spectral) open access dataset for advanced image analysis of spectral radiographic (x-ray) scans, their tomographic reconstruction and the detection of specific materials within such scans. The scans operate at a photon energy range of around 20 keV up to 160 keV (see figure).
The dataset includes — for 2D- as well as 3D spectral data — the corrected (e.g. calibrated) radiographic projections, their tomographic reconstructions (based on 37 projections of 256 detector pixels into a 100×100 pixel CT image per slice) and the corresponding set of segmentation variants demonstrated in Kehl et al. 2018 (preprint).
The dataset is actively curated by the group and new segmentation- and further also classification data will become available in the future.
We hope to contribute with this open access dataset to continuing research on topics of tomographic reconstruction and correction (e.g. compressed-sensing reconstruction, metal artifact reduction, inverse problems in practice), (hyper-)spectral image- and volume segmentation, material classification in CT and — as a final goal of the CIL project — the increase in airport security with these methods.
The MUSIC dataset is part of the CIL (Check-In Luggage) project, funded by the ‘Innovation Fund Denmark’ in collaboration with Innospexion and Exruptive.
The dataset:
MUSIC 2D spectral – HDF5 data (MatLab/Octave/Python): 
MUSIC 2D spectral – MHD data (Python/C++/Paraview/MITK/etc.):
MUSIC 3D spectral – HDF5 data (MatLab/Octave/Python):
MUSIC 3D spectral – MHD data (Python/C++/Paraview/MITK/etc.):
Now, the software for graph-based segmentation given in the paper is also available for Windows 64-bit in binary form:
For building from source – working in Linux and on Windows via MinGW32-x86_64-w64 – you can use the official github link: 