Diffusion kurtosis imaging does not improve differentiation performance of breast lesions in a short clinical protocol

Magn Reson Imaging. 2019 Nov;63:205-216. doi: 10.1016/j.mri.2019.08.007. Epub 2019 Aug 16.

Abstract

Background: Diffusion weighted magnetic resonance imaging (DWI) is known to differentiate between malignant and benign lesions via the apparent diffusion coefficient (ADC). Here, the value of diffusion kurtosis imaging (DKI) for differentiation and further characterization of benign and malignant breast lesions and their subtypes in a clinically feasible protocol is investigated.

Material and methods: This study included 85 patients (with 68 malignant and 73 benign lesions) who underwent 3 T breast DWI using three b values (50, 750, 1500 s/mm2), with a total measurement time < 5 min. ADC maps were calculated from b values 50, 750 s/mm2. The diffusion kurtosis model was fitted to the diffusion weighted images, yielding in each lesion the average kurtosis-corrected diffusion coefficient DK and mean kurtosis K. Histopathology was obtained of radiologically suspicious lesions; follow-up scans were used as a standard of reference for benign appearing lesions. Receiver operating characteristic curves were used to evaluate the parameters' diagnostic performance for differentiation of lesion types and grades. The difference in diffusion parameters between subgroups was analysed statistically using the Wilcoxon rank sum test and Kruskal-Wallis test, applying a Bonferroni correction for multiple testing where necessary.

Results: ADC, DK and K showed significant differences between malignant and benign lesions (p < 10-5). All parameters had similar areas under the curve (AUC) (ADC: 0.92, DK: 0.91, K: 0.89) for differentiation of malignant and benign lesions. Sensitivity was highest for ADC (ADC: 0.96, DK: 0.94, K: 0.93), as well as specificity (ADC: 0.85, DK: 0.82, K: 0.82). ADC and DK showed significant differences between tumor histologic grades (p = 6.8⋅10-4, p = 6.6 · 10-5, respectively), whereas K did not (p = 0.99). All three parameters differed significantly between subtypes of benign lesions (ADC: p < 10-5, DK: p< 10-5, K: p = 4.1·10-4), but not between subtypes of malignant lesions (ADC: p = 0.21, DK: p = 0.25, K: p = 0.08).

Conclusion: DKI parameters and conventional ADC can differentiate between malignant and benign lesions. Differentiation performance was best for ADC. Different tumor grades were significantly different in ADC and DK, which may have an impact on therapy planning and monitoring. In our study, K did not add value to the diagnostic performance of DWI in a clinical setting.

Keywords: Breast neoplasms; Diagnostic imaging; Diffusion kurtosis imaging; Diffusion magnetic resonance imaging; Neoplasms by histologic type.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms
  • Area Under Curve
  • Breast / diagnostic imaging*
  • Breast Neoplasms / diagnostic imaging*
  • Clinical Protocols
  • Diffusion Magnetic Resonance Imaging*
  • Female
  • Humans
  • Image Interpretation, Computer-Assisted / methods*
  • Neoplasm Grading
  • Normal Distribution
  • ROC Curve
  • Reproducibility of Results
  • Sensitivity and Specificity