Quantitative tumor segmentation for evaluation of extent of glioblastoma resection to facilitate multisite clinical trials

Abstract

Standard-of-care therapy for glioblastomas, the most common and aggressive primary adult brain neoplasm, is maximal safe resection, followed by radiation and chemotherapy. Because maximizing resection may be beneficial for these patients, improving tumor extent of resection (EOR) with methods such as intraoperative 5-aminolevulinic acid fluorescence-guided surgery (FGS) is currently under evaluation. However, it is difficult to reproducibly judge EOR in these studies due to the lack of reliable tumor segmentation methods, especially for postoperative magnetic resonance imaging (MRI) scans. Therefore, a reliable, easily distributable segmentation method is needed to permit valid comparison, especially across multiple sites. We report a segmentation method that combines versatile region-of-interest blob generation with automated clustering methods. We applied this to glioblastoma cases undergoing FGS and matched controls to illustrate the method's reliability and accuracy. Agreement and interrater variability between segmentations were assessed using the concordance correlation coefficient, and spatial accuracy was determined using the Dice similarity index and mean Euclidean distance. Fuzzy C-means clustering with three classes was the best performing method, generating volumes with high agreement with manual contouring and high interrater agreement preoperatively and postoperatively. The proposed segmentation method allows tumor volume measurements of contrast-enhanced T 1-weighted images in the unbiased, reproducible fashion necessary for quantifying EOR in multicenter trials.

Figure 1

Preoperative (A and B) and postoperative (C and D) ROI blobs [2-dimensional (2D) and 3D] generated by coarse contouring. Subtraction image (C) accounts for blood product accumulation in resection cavity (dark region). Resultant 2D and 3D tumor segmentations for preoperative (E and F) and postoperative (G and H) contrast-enhanced T₁W images using Fuzzy3 algorithm.

Figure 2

Preoperative and postoperative manual tumor contour volume versus semiautomated segmentation (Otsu4, Fuzzy4, or Fuzzy3 from top to bottom) by two separate readers. CCC, concordance correlation coefficient ± SEM; CI, confidence interval.

Figure 3

Two-dimensional illustration depicting impact of manual and segmented structure overlap on Dice and MED (A) along with mean preoperative and postoperative Dice and MED values with SEM and 95% confidence for Otsu and Fuzzy methods versus manual contouring (B). *P < .05, **P < .05, ⁺P < .05, ⁺⁺P < .05, ^°P < .05, ^°°P < .05, ^aP < .05, and ^aaP < .05. CI, confidence interval.

Figure 4

Preoperative (top) and postoperative (bottom) tumor volumes were generated using Fuzzy3 by reader 1 versus reader 2. CCC, concordance correlation coefficient ± SEM; CI, confidence interval.