Abstract
We sought to investigate, whether texture analysis of diffusional kurtosis imaging (DKI) enhanced by support vector machine (SVM) analysis may provide biomarkers for gliomas staging and detection of the IDH mutation. First-order statistics and texture feature extraction were performed in 37 patients on both conventional (FLAIR) and mean diffusional kurtosis (MDK) images and recursive feature elimination (RFE) methodology based on SVM was employed to select the most discriminative diagnostic biomarkers. The first-order statistics demonstrated significantly lower MDK values in the IDH-mutant tumors. This resulted in 81.1% accuracy (sensitivity = 0.96, specificity = 0.45, AUC 0.59) for IDH mutation diagnosis. There were non-significant differences in average MDK and skewness among the different tumour grades. When texture analysis and SVM were utilized, the grading accuracy achieved by DKI biomarkers was 78.1% (sensitivity 0.77, specificity 0.79, AUC 0.79); the prediction accuracy for IDH mutation reached 83.8% (sensitivity 0.96, specificity 0.55, AUC 0.87). For the IDH mutation task, DKI outperformed significantly the FLAIR imaging. When using selected biomarkers after RFE, the prediction accuracy achieved 83.8% (sensitivity 0.92, specificity 0.64, AUC 0.88). These findings demonstrate the superiority of DKI enhanced by texture analysis and SVM, compared to conventional imaging, for gliomas staging and prediction of IDH mutational status.
Highlights
Exciting advances and an improved understanding of the brain has been facilitated by diffusion-weighted (DWI) MRI, which for brain tumors supplies a measure of tumor cellularity based on the restriction of the free diffusion of water in proliferating tissue[1]
Diffusion kurtosis imaging (DKI) is an attempt to account for this variation and in a more refined approach overcomes this problem by quantifying the deviation from the Gaussian distribution of diffusion properties in brain tissue[3]
The purpose of this study is to validate the previous reports on DKI for IDH mutation status prediction and investigate, whether the diagnostic performance of DKI, either as stand-alone modality or combined with conventional structural imaging, may be enhanced by means of computerized texture analysis assisted by machine learning
Summary
Exciting advances and an improved understanding of the brain has been facilitated by diffusion-weighted (DWI) MRI, which for brain tumors supplies a measure of tumor cellularity based on the restriction of the free diffusion of water in proliferating tissue[1]. The results of this study contradict the previous work by Elkhaled et al.[10] but essentially highlight that (i) the methodology for obtaining DWI parameters in tumor plays a crucial role given the spatial heterogeneity and (ii) the ADC might be not the most appropriate parameter to gauge any altered diffusion properties related to the IDH mutation, and DKI should be employed to improve the diagnostic accuracy[12]. Apart from the DWI studies, Patel et al.[14] made an important contribution by introducing the ‘T2-FLAIR mismatch’ sign as a highly specific morphological feature of the IDH-mutant, 1p/19q non-codeleted molecular subtype of astrocytomas; Park et al.[15] have used the Visually AcceSAble Rembrandt Images (VASARI) library in lower grade gliomas and shown that features like larger proportion of enhancing tissue, multifocal/multicentric distribution, and poorly marginated non-enhancing tumour tissue were independent predictors of an IDH1 wild type tumour. A computer algorithm could be trained on a set of training samples, to discover the most distinctive set of features samples from a large range of potential imaging biomarkers, whilst progressively eliminating non-discriminating lesion patterns in a recursive feature elimination manner
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
