Quantitative multiparametric MRI of ovarian cancer

Abstract

Purpose: To identify parameters associated with ovarian malignancy using multiparametric quantitative magnetic resonance imaging (MRI).

Materials and methods: After Institutional Review Board (IRB) approval, women with ovarian masses underwent preoperative imaging with 3 T MRI. Dynamic contrast-enhanced (DCE)-MRI with pharmacokinetic modeling, quantitative T2 mapping, and diffusion-weighted imaging with quantitative mapping of the water diffusion parameters were performed. Ovarian masses had one or more discreet regions of interest, categorized as cystic or solid, and histologically diagnosed as benign or malignant. Mean region of interest (ROI) values were compared between benign and malignant masses using generalized estimating equations. In addition, we compared classification accuracy for the mean ROI value to a combination of histogram characteristics (standard deviation, skewness, and kurtosis) from T2 map ROIs using logistic regression and ROC curve. The significance level was P = 0.05.

Results: Several DCE-MRI parameters differentiated solid benign from malignant masses. Toft's rate constant (kep ) was significantly higher in malignant masses (P < 0.001), as well as quantitative T2 values (P = 0.003), and signal intensity on T2 weighted imaging (P = 0.008). A linear combination of the mean, standard deviation, skewness, and kurtosis of T2 within solid regions (area under the curve [AUC] 0.90) provided better classification accuracy than the mean of T2 alone (AUC 0.81).

Conclusion: Quantitative parameters from DCE-MRI and T2 mapping can differentiate benign from malignant ovarian masses.

Keywords: investigative techniques; magnetic resonance imaging; methods; neoplasms; ovary.

Figure 1

Example of multiparametric images for a benign serous cystadenoma (subject #37). (a) Anatomical T₂-weighted image with ROI on the large cyst. (b) T₂ map (ms), (c) ADC map (10^-6 mm²/s), and (d) Toft's model k_ep map (min^-1) overlaid on the anatomical image with ROI indicated. The quantitative maps indicate the cystic fluid has a long T₂ (mean 1877 ms), high ADC (mean 2.7 × 10^-3 mm²/s), and no contrast enhancement. Also visible is a hemorrhagic corpus luteum (white arrow), whose ROI was drawn on an adjacent slice, that exhibits short T₂, negligible enhancement, and low SNR on the diffusion images.

Figure 2

Example of multiparametric images for a stage IIIC serous adenocarcinoma of the ovary (subject #21). (a) Anatomical T₂-weighted image with ROI on the largest solid component of the mass. (b) T₂ map (ms), (c) ADC map (10^-6 mm²/s), and (d) Toft's model k_ep map (min^-1) overlaid on the anatomical image with ROI indicated. The maps show this solid region has a relatively short T₂ (118.5 ms) and low ADC (mean 0.86 × 10^-3 mm²/s), suggestive of high cellularity. The high k_ep value (mean 0.45 min^-1) indicates the mass is mostly well-vascularized with regions of poor perfusion. Cystic regions of the same complex mass (white arrows) exhibit variable T₂ values, high ADC, and low enhancement.

Figure 3

Illustration of the distribution of T₂ values within solid ROIs. Top row shows the (a) T₂ map with ROI in a solid region of a benign fibrothecoma (subject #35), and (b) histogram of T₂ values from the ROI. Bottom row shows a (c) T₂ map and ROI in a solid region of a malignant serous adenocarcinoma (subject #33) and (d) associated histogram. In this comparison the benign solid ROI had a shorter mean T₂ value and a more uniform distribution of T₂ values (b) compared to the longer T₂ values and more skewed distribution for the malignant solid ROI (d).