Elsevier

Clinical Imaging

Volume 52, November–December 2018, Pages 44-56
Clinical Imaging

Imaging Physics and Informatics
MR diffusion kurtosis imaging for cancer diagnosis: A meta-analysis of the diagnostic accuracy of quantitative kurtosis value and diffusion coefficient

https://doi.org/10.1016/j.clinimag.2018.06.005Get rights and content

Highlights

  • The first meta-analysis to explore the accuracy of DKI parameters for cancer diagnosis

  • Diagnostic accuracy of D and K were 0.9162 and 0.8908

  • Diagnostic accuracy of D superior to apparent diffusion coefficient

Abstract

Purpose

To perform a meta-analysis for assessing the accuracy of diffusion kurtosis imaging (DKI)-derived quantitative parameters (kurtosis values, K; and corrected diffusion coefficients non-Gaussian bias, D) in separating malignant cancers from benign lesions.

Methods

Relevant studies were searched in PubMed and Cochrane Library databases and were analyzed by Meta-DiSc software.

Results

Fourteen eligible studies involving 1847 lesions in 1107 patients (895 were benign and 952 were malignant) were included. Pooled analysis showed the sensitivity, specificity, positive likelihood ratio (LR), and negative LR were respectively 0.83 (95% CI, 0.79–0.85), 0.83 (95% CI, 0.80–0.86), 4.61 (95% CI, 2.98–7.14), and 0.22 (95% CI, 0.18–0.28) for K, with the overall area under curve (AUC) of 0.89. The sensitivity, specificity, positive LR, and negative LR were 0.85 (95% CI, 0.80–0.88), 0.85 (95% CI, 0.79–0.89), 6.39 (95% CI, 3.14–12.99), and 0.18 (95% CI, 0.14–0.23) for D, with the overall AUC of 0.92. The sensitivity, specificity, positive LR, and negative LR for apparent diffusion coefficient (ADC) derived from standard diffusion-weighted imaging (DWI) were 0.82 (95% CI, 0.79–0.84), 0.85 (95% CI, 0.82–0.88), 4.75 (95% CI, 3.38–6.68), and 0.24 (95% CI, 0.19–0.29), with the overall AUC of 0.89. The superiority of D to K and ADC was also confirmed by the subgroup analysis of prostate cancer.

Conclusion

Our findings suggest that DKI should be added to the routine imaging protocol for screening cancer, with the highest diagnostic accuracy of diffusion coefficients.

Introduction

Diffusion-weighted imaging (DWI) is a widely used magnetic resonance imaging (MRI) technique type for the detection of cancer from the benign lesions. DWI uses non-contrast enhanced MRI sequence to measure the Brownian motion of water molecules in biological tissues [1]. The diffusivity of water can be quantified by calculating the apparent diffusion coefficient (ADC) values. The relatively high cellularity in malignant tumors commonly leads to a diffusion restriction and thus induces low ADC values yielded [2, 3]. Accordingly, quantitative ADC values could potentially serve as an effective biomarker for non-invasive diagnosis and prognosis of cancer. Unfortunately, some studies indicate there is a similar ADC value between benign and malignant lesions (i.e. idiopathic granulomatous mastitis vs malignant breast lesions [4]; ameloblastomas vs simple bone cysts [5]). This may be attributed to the assumption that the diffusing movement of free water molecules follows Gaussian distribution in the conventional DWI model. In fact, diffusion of water molecules can also be restricted by the presence of various microstructural barriers in biological tissues, such as cellular compartments (intracellular and extracellular spaces) and cell membranes [6], which may make the water molecules diffuse in a non-Gaussian model [7]. Hereby, more advanced diffusion MR imaging techniques that measure diffusional non-Gaussianity are highly required to further assist diagnosis of cancer from the benign lesions.

Diffusion kurtosis imaging (DKI) is an extension of diffusion tensor imaging that assesses the microstructure properties of tissues in a non-Gaussian diffusion-weighted model. From this model, two quantitative parameters can be calculated, including kurtosis values (K, representing deviation from a Gaussian distribution) and diffusion coefficient (D, defining as a corrected ADC for non-Gaussian bias) [7]. Theoretically, DKI with higher K and lower D may exhibit a substantially higher sensitivity than conventional DWI with ADC calculation for cancer detection [8], which has been demonstrated by several studies as following: Jiang et al. demonstrated both K and D had a significantly higher accuracy compared with ADC for the differentiation between benign and malignant sinonasal lesions [sensitivity, 95.70% and 82.60% vs 69.60%; the area under curve (AUC), 0.88 and 0.84 vs. 0.76] [9]. The study of Sun et al. also showed a significantly higher specificity for the differentiation of malignant from benign lesions with the use of K and D than with the use of the ADC (83.00% and 83.00% vs 76.00%) [10]. However, the results of currently published studies on this topic remain controversial. Das et al. found the diagnostic accuracy and AUC of D were not significantly higher in differentiating malignancies from benign pulmonary nodules as compared to ADC (accuracy: 85.70% vs 77.14%; AUC: 0.87 vs 0.81) [11]. Also, Tamura et al. observed no significant difference in AUC between K and ADC for diagnosis of prostate cancer [12]. Therefore, the purpose of this study was to further evaluate the diagnostic performance of quantitative K and D for separating malignant cancer from the benign lesions by performing a systematic meta-analysis, which has not been reported [13].

Section snippets

Search strategy and inclusion criteria

This meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [14].

A comprehensive literature search was performed on the PubMed and the Cochrane Library databases to obtain all relevant studies published in English before December of 2017, but without a lower date limit. The following keywords were used: (diffusion kurtosis imaging or diffusion kurtosis) and (cancer or lesions or carcinoma). In addition, all

Characteristics of eligible studies

The results of literature search strategy are illustrated in the PRISMA flowchart (Fig. 1). By searching with the predefined key words, 559 publications were yielded, among which 364 studies were excluded due to duplication. After reviewing the article titles, abstracts and full text, 14 manuscripts ultimately fulfilled the inclusion and exclusion criteria and were selected for data extraction. Table 1 presents the main characteristics of the included studies. These 14 eligible studies were

Discussion

In present study, we, for the first time, used a meta-analysis to investigate the diagnostic accuracy of DKI-derived quantitative parameters (K and D) in the discrimination between malignant cancer and benign lesions. Pooled results indicated both K and D had a good or excellent diagnostic performance in separating malignant cancer from benign lesions, but D may be more superior because it had the higher AUC (0.92 ± 0.02 vs 0.89 ± 0.01) and only its positive LR was >5.0 (6.39, 95%

Funding

This study was supported by National Natural Science Foundation of China (No. 81671679); Shanghai Municipal Science and Technology Development Found (No. 15411952000); Shanghai Shenkang Development Center Found (No. SHDC12014227).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Acknowledgments

None.

References (33)

  • C.E. Metz

    Basic principles of ROC analysis

    Semin Nucl Med

    (1978)
  • H. Bickel et al.

    Diffusion-weighted imaging of breast lesions: region-of-interest placement and different ADC parameters influence apparent diffusion coefficient values

    Eur Radiol

    (2017)
  • M.A.H.E. Bakry et al.

    Role of diffusion weighted imaging and dynamic contrast enhanced magnetic resonance imaging in breast tumors

    Egypt J Radiol Nucl Med

    (2015)
  • R.F. Qu et al.

    Differential diagnosis of benign and malignant breast tumors using apparent diffusion coefficient value measured through diffusion-weighted magnetic resonance imaging

    J Comput Assist Tomogr

    (2015)
  • R. Yilmaz et al.

    Magnetic resonance imaging features of idiopathic granulomatous mastitis: is there any contribution of diffusion-weighted imaging in the differential diagnosis?

    Radiol Med

    (2016)
  • S. Eida et al.

    Apparent diffusion coefficient-based differentiation of cystic lesions of the mandible

    Oral Radiol

    (2012)
  • R.J. Gillies et al.

    MRI of the tumor microenvironment

    J Magn Reson Imaging

    (2002)
  • J.H. Jensen et al.

    Diffusional kurtosis imaging: the quantification of non-Gaussian water diffusion by means of magnetic resonance imaging

    Magn Reson Med

    (2005)
  • J. Budjan et al.

    Diffusion kurtosis imaging of the liver at 3 Tesla: in vivo comparison to standard diffusion-weighted imaging

    Acta Radiol

    (2018)
  • J.X. Jiang et al.

    Diffusion kurtosis imaging for differentiating between the benign and malignant sinonasal lesions

    J Magn Reson Imaging

    (2016)
  • K. Sun et al.

    Breast cancer: diffusion kurtosis MR imaging-diagnostic accuracy and correlation with clinical-pathologic factors

    Radiology

    (2015)
  • S.K. Das et al.

    Non-Gaussian diffusion imaging for malignant and benign pulmonary nodule differentiation: a preliminary study

    Acta Radiol

    (2017)
  • C.S.H. Tamura et al.

    Diffusion kurtosis imaging study of prostate cancer: preliminary findings

    J Magn Reson Imaging

    (2014)
  • A.D. Falk et al.

    Glioma grade discrimination with MR diffusion kurtosis imaging: a meta-analysis of diagnostic accuracy

    Radiology

    (2017)
  • L.A. Stewart et al.

    Preferred reporting items for systematic review and meta-analyses of individual participant data: the PRISMA-IPD statement

    JAMA

    (2015)
  • P.F. Whiting et al.

    QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies

    Ann Intern Med

    (2011)
  • Cited by (12)

    • Detecting the muscle invasiveness of bladder cancer: An application of diffusion kurtosis imaging and tumor contact length

      2022, European Journal of Radiology
      Citation Excerpt :

      Diffusion kurtosis imaging (DKI), derived from the principle of DWI, was first described by Jensen et al[11]. Many studies have contributed tremendously to understanding the relationship between DKI and malignancy, pathological grade, and clinical stage of tumors in whole body[12–15]. However, only a few articles have probed the application of DKI in bladder cancer[16–18].

    • Optimal b-values for diffusion kurtosis imaging of the liver and pancreas in MR examinations

      2019, Physica Medica
      Citation Excerpt :

      For that reason, it is necessary to standardise the imaging protocol. In the literature, some studies reported different numbers of b-values and distinct diffusion gradient directions required for the DKI sequence [3,4,8,15,18–21]. Studies focused on abdominal organs using three to thirteen b-values.

    View all citing articles on Scopus
    View full text