A Multiparametric Model for Mapping Cellularity in Glioblastoma Using Radiographically Localized Biopsies

Abstract

Background and purpose: The complex MR imaging appearance of glioblastoma is a function of underlying histopathologic heterogeneity. A better understanding of these correlations, particularly the influence of infiltrating glioma cells and vasogenic edema on T2 and diffusivity signal in nonenhancing areas, has important implications in the management of these patients. With localized biopsies, the objective of this study was to generate a model capable of predicting cellularity at each voxel within an entire tumor volume as a function of signal intensity, thus providing a means of quantifying tumor infiltration into surrounding brain tissue.

Materials and methods: Ninety-one localized biopsies were obtained from 36 patients with glioblastoma. Signal intensities corresponding to these samples were derived from T1-postcontrast subtraction, T2-FLAIR, and ADC sequences by using an automated coregistration algorithm. Cell density was calculated for each specimen by using an automated cell-counting algorithm. Signal intensity was plotted against cell density for each MR image.

Results: T2-FLAIR (r = -0.61) and ADC (r = -0.63) sequences were inversely correlated with cell density. T1-postcontrast (r = 0.69) subtraction was directly correlated with cell density. Combining these relationships yielded a multiparametric model with improved correlation (r = 0.74), suggesting that each sequence offers different and complementary information.

Conclusions: Using localized biopsies, we have generated a model that illustrates a quantitative and significant relationship between MR signal and cell density. Projecting this relationship over the entire tumor volume allows mapping of the intratumoral heterogeneity in both the contrast-enhancing tumor core and nonenhancing margins of glioblastoma and may be used to guide extended surgical resection, localized biopsies, and radiation field mapping.

Fig 1.

Whole-cell counting. A, Digitized, low-power magnification view of a single H&E-stained slide. Two representative 400× fields from this single tissue specimen of relatively lower (B) and higher (C) cell density illustrate the tissue heterogeneity present at a microscopic level. Stained cellular nuclei identified by the automated counting algorithm are outlined in green. D, The “heat map” demonstrates distribution of cell density at the level of a HPF throughout the tissue sample.

Fig 2.

Cell-counting statistics. A, Comparison between manual and automated cell counts for 25 high-power fields of various cellular densities. Correlation is high (r = 0.984), suggesting that the automated algorithm accurately reflects manual counts. B, For each biopsy sample, the median cell density of all HPFs is compared with that of the 98th percentile. A relatively strong linear correlation is preserved (r = 0.901), suggesting that the 98th percentile cell density simply represents a linear translation of the median cell density. C, Correlation analysis is repeated for all percentiles (0–100). With the exception of extreme values, most percentiles retain a strong linear correlation (r > 90%) with the median cell density.

Fig 3.

Cell count versus MR signal intensity. Scatterplots demonstrate median cell density as a function of signal intensity on ADC (A), T2-FLAIR (B), and T1-postcontrast subtraction sequences (C) correlated by using single-variable regression analysis. The linear regression and Pearson correlation (r) were significant (P < .05) for all 3 sequences. D, The scatterplot shows the actual and predicted cell counts as estimated by combining all 3 imaging modalities in a multiple-variable regression model.

Fig 4.

Correlation versus distance from the biopsy. Scatterplots demonstrate the correlation between cell density and signal intensity for each MR image (ADC, T2-FLAIR, T1-postcontrast subtraction) obtained by taking the mean of concentric spheric shells of voxels at an increasing distance from the original biopsy point. Notably, the correlations drop to 0 at a radius of approximately 5 mm (∼10 voxels), providing an estimate of the spatial accuracy of the biopsy location. T1SUB indicates T1-subtraction.

Fig 5.

Whole-tumor model overlay. Estimated cellularity by applying the multiple regression model on a voxelwise basis across the tumor. The model is derived from linear regression by using ADC, T2-FLAIR, and T1-postcontrast sequences shown in the inset on the left. In the right panels, corresponding biopsy specimens (400× magnification, H&E stained sections) are shown from 2 regions obtained on the same section, highlighting the considerable variation in cellularity in and around the region of contrast enhancement (demarcated by a white outline).