Representation of brain tumours in two-dimensional diffusion kurtosis imaging (DKI) histograms

Abstract

Malignant brain tumours often possess an unfavourable prognosis and can severely diminish quality of life. Therefore, it is vital to improve the diagnostic process so that the differentiation between various tumour subgroups can be facilitated. The objective of this work was to analyse two-dimensional histograms, which were generated with the help of diffusion kurtosis imaging (DKI). For this purpose, the DKI histograms of 10 subjects without a brain tumour and 39 patients with a brain tumour were analysed. The aim was to investigate, whether any possible differences could be recognised between these two groups and to compare the various tumour subcategories gliomas, meningiomas and brain metastases. Regarding similarities between healthy subjects and patients, common histogram areas were identified that corresponded to specific regions like the ventricles or the subarachnoid space in the majority of subjects. Moreover, highlighting a specific histogram area marked corresponding layer-like regions in the brain. In tumour patients, these layers usually did not extend into the pathological tissue, rendering it distinguishable from physiological brain tissue. With respect to the analysed brain tumours, gliomas and their peritumoral oedema were found to correspond to a specific spike-like extension of the histogram, which was not observed when analysing a healthy brain. This spike also represented the oedema from meningiomas and metastases and can therefore be referred to as the glioma/oedema spike. When comparing oligodendrogliomas and astrocytomas with glioblastomas, it was possible to recognize a slightly different shape and orientation of the glioma/oedema histogram extension. In addition, the data suggest that various tumour subgroups of meningiomas and metastases may be distinguished. Moreover, it was mostly possible to differentiate between these two tumour categories and gliomas. In the future, these observations about the different behaviour of the tumour groups could improve the diagnostic process, for example with the help of artificial intelligence. Besides, the DKI histograms could be used for automatic tumour segmentation.