Comparison of unsupervised classification methods for brain tumor segmentation using multi-parametric MRI.

Tumor segmentation is a particularly challenging task in high-grade gliomas (HGGs), as they are among the most heterogeneous tumors in oncology. An accurate delineation of the lesion and its main subcomponents contributes to optimal treatment planning, prognosis and follow-up. Conventional MRI (cMRI) is the imaging modality of choice for manual segmentation, and is also considered in the vast majority of automated segmentation studies. Advanced MRI modalities such as perfusion-weighted imaging (PWI), diffusion-weighted imaging (DWI) and magnetic resonance spectroscopic imaging (MRSI) have already shown their added value in tumor tissue characterization, hence there have been recent suggestions of combining different MRI modalities into a multi-parametric MRI (MP-MRI) approach for brain tumor segmentation. In this paper, we compare the performance of several unsupervised classification methods for HGG segmentation based on MP-MRI data including cMRI, DWI, MRSI and PWI. Two independent MP-MRI datasets with a different acquisition protocol were available from different hospitals. We demonstrate that a hierarchical non-negative matrix factorization variant which was previously introduced for MP-MRI tumor segmentation gives the best performance in terms of mean Dice-scores for the pathologic tissue classes on both datasets.

White matter organization in relation to upper limb motor control in healthy subjects: exploring the added value of diffusion kurtosis imaging.

Diffusion tensor imaging (DTI) characterizes white matter (WM) microstructure. In many brain regions, however, the assumption that the diffusion probability distribution is Gaussian may be invalid, even at low b values. Recently, diffusion kurtosis imaging (DKI) was suggested to more accurately estimate this distribution. We explored the added value of DKI in studying the relation between WM microstructure and upper limb coordination in healthy controls (N = 24). Performance on a complex bimanual tracking task was studied with respect to the conventional DTI measures (DKI or DTI derived) and kurtosis metrics of WM tracts/regions carrying efferent (motor) output from the brain, corpus callosum (CC) substructures and whole brain WM. For both estimation models, motor performance was associated with fractional anisotropy (FA) of the CC-genu, CC-body, the anterior limb of the internal capsule, and whole brain WM (r s range 0.42-0.63). Although DKI revealed higher mean, radial and axial diffusivity and lower FA than DTI (p < 0.001), the correlation coefficients were comparable. Finally, better motor performance was associated with increased mean and radial kurtosis and kurtosis anisotropy (r s range 0.43-0.55). In conclusion, DKI provided additional information, but did not show increased sensitivity to detect relations between WM microstructure and bimanual performance in healthy controls.

Automated extraction of the femoral anatomical axis for determining the intramedullary rod parameters in total knee arthroplasty.

The automated extraction of anatomical reference parameters may improve speed, precision and accuracy of surgical procedures. In this study, an automated method for extracting the femoral anatomical axis (FAA) from a 3D surface mesh, based on geometrical entity fitting, is presented. This was applied to conventional total knee arthroplasty, which uses an intramedullary rod (FIR) to orient the femoral prosthesis with respect to the FAA. The orientation and entry point of a FIR with a length of 200 mm are automatically determined from the FAA, as it has been shown that errors in these parameters may lead to malalignment of the mechanical axis. Moreover, the effect of partially scanning the leg was investigated by creating reduced femur models and comparing the results with the full models. Precise measurements are obtained for 50 models by using a central and two outer parts, with lengths of 20 and 120 mm, which correspond to 58% of the mean femoral length. The deviations were less than 2 mm for the FAA, 2.8 mm for the FAA endpoints and 0.7° and 1.3 mm for the FIR orientation and entry point. The computer-based techniques might eventually be used for preoperative planning of total knee arthroplasty.

A new method for improved standardisation in three-dimensional computed tomography cephalometry.

Interest for three-dimensional computed tomography cephalometry has risen over the last two decades. Current methods commonly rely on the examiner to manually point-pick the landmarks and/or orientate the skull. In this study, a new approach is presented, in which landmarks are calculated after selection of the landmark region on a triangular model and in which the skull is automatically orientated in a standardised way. Two examiners each performed five analyses on three skull models. Landmark reproducibility was tested by calculating the standard deviation for each observer and the difference between the mean values of both observers. The variation can be limited to 0.1 mm for most landmarks. However, some landmarks perform less well and require further investigation. With the proposed reference system, a symmetrical orientation of the skulls is obtained. The presented methods contribute to standardisation in cephalometry and could therefore allow improved comparison of patient data.