Apparent diffusion coefficients measured using different regions of interest in differentiating borderline from malignant ovarian tumors

Abstract

Background

Apparent diffusion coefficients (ADCs) measured using different regions of interest (ROIs) are widely used in differentiating ovarian tumors.

Purpose

To evaluate the diagnostic performance of ADCs with different ROIs in differentiating borderline ovarian tumors (BOTs) from malignant ovarian tumors (MOTs).

Material and Methods

Thirty-five BOTs and 54 MOTs who underwent diffusion-weighted magnetic resonance imaging (MRI) were evaluated retrospectively. ADC values were independently measured using five ROI methods: round; rectangle; hot-spot; edge drawing; and five sample ROIs. The inter- and intraclass correlation coefficients (ICCs), one-way analysis of variance, receiver operating characteristic curve analysis, and unpaired t-tests were used to perform the statistical analyses.

Results

The measurement reproducibility of the minimum ADC and mean ADC values were good or excellent for BOTs and MOTs (ICC = 0.70–0.95). The minimum and mean ADC value by the edge drawing ROI were significantly higher than those of the other ROI methods (both P < 0.05). The area under the curve (AUC) of the minimum ADC value was less than that of the mean ADC value from the five ROI methods, whereas the AUCs of the mean ADC values from the round ROI and five sample ROIs were significantly larger than those of the other ROI methods (P < 0.05). The minimum and mean ADC values from the five ROI methods showed significant differences between BOTs and MOTs (all P < 0.05).

Conclusion

The ROI shape influences the diagnostic performance of ADC value for differentiating BOTs from MOTs. The mean ADC values from the round ROI and five sample ROIs have better diagnostic efficiency.