Glioma grade discrimination with MR diffusion kurtosis imaging: A meta-Analysis of diagnostic accuracy

Forskningsoutput: TidskriftsbidragArtikel i vetenskaplig tidskrift

Abstract

Purpose: To assess the diagnostic test accuracy and sources of heterogeneity for the discriminative potential of diffusion kurtosis imaging (DKI) to differentiate low-grade glioma (LGG) (World Health Organization [WHO] grade II) from high-grade glioma (HGG) (WHO grade III or IV). Materials and Methods: The Cochrane Library, Embase, Medline, and the Web of Science Core Collection were systematically searched by two librarians. Retrieved hits were screened for inclusion and were evaluated with the revised tool for quality assessment for diagnostic accuracy studies (commonly known as QUADAS-2) by two researchers. Statistical analysis comprised a random-effects model with associated heterogeneity analysis for mean differences in mean kurtosis (MK) in patients with LGG or HGG. A bivariate restricted maximum likelihood estimation method was used to describe the summary receiver operating characteristics curve and bivariate meta-regression. Results: Ten studies involving 430 patients were included. The mean difference in MK between LGG and HGG was 0.17 (95% confidence interval [CI]: 0.11, 0.22) with a z score equal to 5.86 (P < .001). The statistical heterogeneity was explained by glioma subtype, echo time, and the proportion of recurrent glioma versus primary glioma. The pooled area under the curve was 0.94 for discrimination of HGG from LGG, with 0.85 (95% CI: 0.74, 0.92) sensitivity and 0.92 (95% CI: 0.81, 0.96) specificity. Heterogeneity was driven by neuropathologic subtype and DKI technique. Conclusion: MK shows high diagnostic accuracy in the discrimination of LGG from HGG.

Detaljer

Författare
Enheter & grupper
Externa organisationer
  • Karolinska Institute
  • Karolinska University Hospital
  • Uppsala universitet, Historiska institutionen
Forskningsområden

Ämnesklassifikation (UKÄ) – OBLIGATORISK

  • Radiologi och bildbehandling
Originalspråkengelska
Sidor (från-till)119-127
Antal sidor9
TidskriftRadiology
Volym287
Utgivningsnummer1
StatusPublished - 2018 apr 1
PublikationskategoriForskning
Peer review utfördJa