Breast cancer recurrence risk prediction using whole-lesion histogram analysis with diffusion kurtosis imaging

Clin Radiol. 2020 Mar;75(3):239.e1-239.e8. doi: 10.1016/j.crad.2019.10.015. Epub 2019 Nov 22.

Abstract

Aim: To explore the role of whole-lesion histogram analysis on diffusion kurtosis imaging (DKI) for predicting breast cancer 21-gene expression profiles and recurrence scores (RSs).

Materials and methods: This retrospective study was approved by the institutional review board, and informed consent was waived. Seventy-two patients with breast cancer, who underwent genomic testing and DKI (b values: 0-2,800 s/mm2) were enrolled. Patients were divided into low-, intermediate-, and high-RS groups based on their genomic testing results. Diffusivity (D), kurtosis (K), total apparent diffusion coefficient (Total ADC), and ADC0-700 histogram parameters were calculated. Student's t-test, Wilcoxon signed-rank test, Jonckheere-Terpstra test, receiver operating characteristic curves, and Spearman's correlation were used for the statistical analysis.

Results: Total ADC mean/30%/50%/70%, D mean/50%, K mean/30%/50%/70% showed significant differences among the low-, intermediate-, and high-RS groups (p ≤ 0.001, respectively). K50% had the strongest correlation with RSs (correlation coefficient, CC: 0.55). Furthermore, K50% was also correlated with the expression of gene PR, BCL2 and CEGP1 (CC: 0.45, -0.41, -0.41).

Conclusions: Whole-lesion histogram analysis of DKI parameters can be a useful tool for RS prediction of breast cancer. K50% was found to be the most promising parameter for RS prediction.

MeSH terms

  • Adult
  • Aged
  • Breast Neoplasms / diagnostic imaging*
  • Breast Neoplasms / genetics
  • Breast Neoplasms / pathology
  • Diffusion Magnetic Resonance Imaging / methods*
  • Female
  • Gene Expression Profiling
  • Humans
  • Image Interpretation, Computer-Assisted
  • Middle Aged
  • Neoplasm Recurrence, Local / diagnostic imaging*
  • Neoplasm Recurrence, Local / genetics
  • Neoplasm Recurrence, Local / pathology
  • Predictive Value of Tests
  • Retrospective Studies
  • Risk
  • Sensitivity and Specificity