Evaluation of vascular permeability in gliomas by using parameter K2 from dynamic susceptibility contrast data-sets and histogram analysis.
Toshiaki Taoka1, Hisashi Kawai1, Toshiki Nakane1, Toshiteru Miyasaka2, and Shinji Naganawa1

1Radiology, Nagoya University, Nagoya, Japan, 2Radiology, Nara Medical University, Kashihara, Japan

Synopsis

Permeability images can provide additional information to perfusion images in the clinical practice of brain tumors. However, permeability imaging by dynamic contrast enhancement methods requires a long acquisition time. K2 is an index that represents permeability and can be calculated from the dataset of perfusion images with the dynamic susceptibility contrast method, which requires a short acquisition time. In the current study, we calculated K2 for various grades of gliomas and found that K2 showed a significantly higher 20th percentile value in Grade IV compared to Grade III gliomas, providing useful information for grading of gliomas.

Purpose

Permeability imaging by the dynamic contrast enhancement (DCE) method can provide Ktrans, which is useful in glioma grading. However, DCE methods require long imaging times of up to 10 minutes. Tissue permeability can also be calculated using the dynamic susceptibility contrast (DSC) method, which can be performed in the first pass time window. K2 is an index that represents permeability and can be calculated from the dataset obtained with the DSC method. The purpose of the current study is to evaluate the usefulness of K2 obtained with the DSC method for grading of gliomas, as well as relative cerebral blood volume (rCBV) from the same datasets. We also measured DCE to calculate Ktrans and evaluate the correlation with K2.

Methods

[Cases] Twenty-two patients with glioma (Grade II: 5, III: 6, IV: 11) underwent DSC studies, including eight patients in which both DSC and DCE studies were performed within 10 days, but on different days. [Imaging] Imaging was performed with 3-T magnetic resonance imaging (MRI) (Siemens MAGNETOM Verio). DSC perfusion datasets were acquired using the EPI sequence (TR/TE = 1370/35 ms), and DCE permeability datasets were acquired using the gradient echo sequence (TR/TE = 3.88/1.31 ms, FA = 12 deg). [Post-processing] K2, rCBV, and Ktrans were calculated using post-processing software Olea Sphere v2.3 (Olea Medical, La Ciotat, France). Ktrans computations were made with the Extended Tofts method [Ref. 1], and K2 and rCBV computations with leakage correction were made with the Boxerman method [Ref. 2]. [Analysis] We placed regions of interest (ROIs) on the tumor based on Response Assessment in Neuro-Oncology WG (RANO) criteria [Ref. 3]. Thus, for a "measurable tumor" with contrast enhancing lesions with clearly defined margins, we placed the ROI on contrast enhancing MRI, and for a "non-measurable tumor" without clearly defined margins on contrast MRI, we placed the ROI on FLAIR (Fig. 1). We performed histogram analysis of ROIs of the tumors and acquired 20th percentile values for leakage-corrected CBV (CBV20%ile), K2 (K220%ile), and for cases with a DCE study, Ktrans (Ktrans20%ile). We analyzed (1) the correlation between K220%ile and Ktrans20%ile and (2) the statistical difference between CBV20%ile and K220%ile values among glioma grades (II, III, and IV). (3) We also performed the same post-processing for the 5%ile, 10%ile, 15%ile, 30%ile, and the mean values of the same ROIs and performed receiver-operating characteristic (ROC) analysis for grading gliomas (II vs. III and III vs. IV) by rCBV and K2.

Results

(1) We found a statistically significant correlation between K2 and Ktrans (r = 0.717, p < 0.05, Fig. 2). (2) CBV was significantly different between Grades II and III, and between Grades II and IV. K2 was significantly different (p < 0.05) between Grades II and IV, and between Grades III and IV (Fig. 3). (3) ROC analysis (Fig. 4) indicated that the area under the curve for discrimination between Grades II and III by CBV was: 5%ile: 0.83, 10%ile: 0.77, 15%ile: 0.87, 20%ile: 0.87, 30%ile: 0.87, mean value: 0.87. For K2, analysis showed: 5%ile: 0.53, 10%ile: 0.60, 15%ile: 0.67, 20%ile: 0.70, 30%ile: 0.78, mean value: 0.73. Similarly, discrimination between Grades III and IV by CBV was: 5%ile: 0.67, 10%ile: 0.59, 15%ile: 0.59, 20%ile: 0.53, 30%ile: 0.58, mean value: 0.53. For K2, analysis showed: 5%ile: 0.89 10%ile: 0.91, 15%ile: 0.94, 20%ile: 0.91, 30%ile: 0.86, mean value: 0.32. (Underlined values are >0.8.)

Discussion

K2 is a byproduct of the mathematical leakage correction process and refers to the leakage rate detected in the DSC method. Although quantitative interpretation of K2 is complex, it is proportional to the vascular permeability. Here, Ktrans with the DCE method and K2 with the DSC method correlated well when compared using histogram analysis. The CBV value with the DSC method showed a statistically significant difference between Grades II and III, both in mean value and with histogram analysis. In contrast, the K2 value with the DSC method showed a statistically significant difference between Grades III and IV only in histogram analysis. Thus, the K2 value obtained with the DSC method can be used as a substitute for Ktrans with the DCE method when histogram analysis (~30%ile) is performed.

Conclusion

The K2 value calculated from the DSC dataset, which can be obtained with a short acquisition time, was correlated with Ktrans obtained with the DCE method and seems useful for glioma grading, especially for discrimination between Grades III and IV when evaluated using histogram analysis.

Acknowledgements

No acknowledgement found.

References

1. Tofts PS et al. J Magn Reson Imaging. 1999; 10:223-32.

2. Boxerman JL et al. AJNR 2006; 27:859-67.

3. Wen PY et al. J Clin Oncol 2010; 28:1963-1972.

Figures

CBV images and K2 images. ROIs were placed by RANO criteria.

Correlation between K2 20%ile by DSC method and Ktrans20%ile by DCE method.

Statistical difference of CBV20%ile and K220%ile values among grades (II, III and IV) of gliomas.

ROC analysis for grading gliomas (II vs. III and III vs. IV) by rCBV and K2.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1381