Type:
Educational Exhibit
Keywords:
Neuroradiology brain, Liver, MR-Spectroscopy, Computer Applications-General, Cancer, Education and training
Authors:
D. Sima, D. Smeets, D. Loeckx; Leuven/BE
DOI:
10.1594/ecr2013/C-0689
Imaging findings OR Procedure details
Spatial resolution of maps derived from MRSI data can be increased artificially by mathematical interpolation.
As a first possibility,
standard 2D interpolation can be performed on each of the metabolite maps of interest.
For instance,
the resolution of the metabolic maps in Figure 1 could be artificially increased such that it matches the MRI resolution.
This can be performed by traditional mathematical tools such as bilinear interpolation or bicubic spline interpolation.
Although this could lead to a visual enhancement of the metabolite maps,
it adds no relevant information and no fine details to the maps.
As a second possibility,
standard mathematical interpolation could be applied on the spectra in the 2D MRSI grid instead of on the metabolite maps.
This means that new spectra would be artificially created at a spatial resolution higher than that of the original MRSI grid.
Such a method has been applied for the results in Fig.
3,
where the anatomic MRI has first been segmented into gray matter,
white matter,
cerebrospinal fluid and tumour tissue; then,
for all pixels in the tumour region,
the corresponding spectra were selected from a high-resolution interpolated MRSI grid.
An automatic classification method (trained on a large number of example spectra from various brain tumors) has then been applied in order to detect the tumor type and grade.
Finally,
the segmentation and classification based on MRI and MRSI have been visually presented as a nosologic image (see Fig.
3).
We expect that further improvements in MRSI-based quantification and classification can be obtained if the spatial resolution is increased using adaptive interpolation methods that can take into account anatomic information instead of general purpose mathematical interpolation.
As a task in the European project TRANSACT (2013-2017),
MRSI and anatomical MR images will be combined with the goal of up-sampling the spectroscopic image.
The proposed approach will rely on exploiting all available spatial prior knowledge,
such as anatomic structures and,
possibly,
diffusion and perfusion information from the tissue.
Anatomical and metabolic images at higher resolution will then be used to quantify more accurately the evolution of tumour size and shape in brain tumor patients,
and monitor therapy progress and tumour reoccurrence after treatment.
Additionally,
we will apply the spatial enhancement on MRSI acquired from patients with neurological disorders in order to detect metabolite content in a more sensitive way.