Reproducibility of Dynamic Contrast-enhanced MR Imaging. Part II. Comparison of Intra- and Interobserver Variability with Manual Region of Interest Placement versus Semiautomatic Lesion Segmentation and Histogram Analysis

Abstract

A semiautomatic lesion segmentation and histogram analysis approach can provide a significant reduction in interobserver variability for DCE MR imaging measurements of K(trans) when compared with manual ROI methods, whereas intraobserver reproducibility is improved to some extent.

Differentiation between Glioblastoma and Primary Central Nervous System Lymphoma Using Dynamic Susceptibility Contrast-Enhanced Perfusion MR Imaging: Comparison Study of the Manual versus Semiautomatic Segmentation Method

“…To the best of our knowledge, the diagnostic accuracy of the nCBV obtained from the random manual ROIs placement and the semiautomatic segmentation method in brain tumor differential diagnosis has not been compared yet. Prior research assessed the manual and semiautomatic approach in the mass evaluation of various organs, but most of them focused on the reproducibility issue (22,23,32). The semiautomatic lesion segmentation significantly improves both intraobserver and interobserver reproducibility in the GBM nCBV measurement (22) and reduces interobserver variability in the permeability parameter (Ktrans) measurement from dynamic contrast enhancement MRI in uterine fibroid (23).…”

“…Prior research assessed the manual and semiautomatic approach in the mass evaluation of various organs, but most of them focused on the reproducibility issue (22,23,32). The semiautomatic lesion segmentation significantly improves both intraobserver and interobserver reproducibility in the GBM nCBV measurement (22) and reduces interobserver variability in the permeability parameter (Ktrans) measurement from dynamic contrast enhancement MRI in uterine fibroid (23). Bauknecht et al (32) showed improved reproducibility in the semiautomatic volume measurement of brain metastasis compared to the conventional linear method.…”

“…Although the diagnostic value of nCBV in brain tumor evaluations is quite established, the method to obtain the nCBV is not standardized (6). The conventional hot-spot method determines a few small ROIs manually which make the process considerably observer dependent and susceptible to sampling bias (23,25). Because GBM has a highly variable nCBV from pixel to pixel and a broad range of nCBV values in its solid area (14,(24)(25)(26)33), the nCBV of GBM is more likely to fluctuate even with the small variation of ROIs.…”

“…The reproducibility and the inter-and intra-observer variation of the manual placement of ROI have been issues, and several studies have suggested that automatic or semiautomatic segmentation methods reduce observerrelated problems (22)(23)(24)(25). However, whether the softwareassisted method shows better diagnostic performance than manual method has not been validated enough.…”

“…Because the range of the rCBV value in the solid area of GBM is wide (25,26), the sampling bias from the random manual ROI selection can significantly affect the accuracy of the differential diagnosis of GBM from PCNSL based on the rCBV. Therefore, we hypothesized that the semiautomatic segmentation method for the tumoral rCBV measurement can improve the preoperative diagnostic accuracy of the differentiation between GBM and PCNSL, because of its better reproducibility and observer-independence compared to random manual ROI selection (22,23).…”