Morphometic analysis of TCGA glioblastoma multiforme

Abstract

Background: Our goals are to develop a computational histopathology pipeline for characterizing tumor types that are being generated by The Cancer Genome Atlas (TCGA) for genomic association. TCGA is a national collaborative program where different tumor types are being collected, and each tumor is being characterized using a variety of genome-wide platforms. Here, we have developed a tumor-centric analytical pipeline to process tissue sections stained with hematoxylin and eosin (H&E) for visualization and cell-by-cell quantitative analysis. Thus far, analysis is limited to Glioblastoma Multiforme (GBM) and kidney renal clear cell carcinoma tissue sections. The final results are being distributed for subtyping and linking the histology sections to the genomic data.

Results: A computational pipeline has been designed to continuously update a local image database, with limited clinical information, from an NIH repository. Each image is partitioned into blocks, where each cell in the block is characterized through a multidimensional representation (e.g., nuclear size, cellularity). A subset of morphometric indices, representing potential underlying biological processes, can then be selected for subtyping and genomic association. Simultaneously, these subtypes can also be predictive of the outcome as a result of clinical treatments. Using the cellularity index and nuclear size, the computational pipeline has revealed five subtypes, and one subtype, corresponding to the extreme high cellularity, has shown to be a predictor of survival as a result of a more aggressive therapeutic regime. Further association of this subtype with the corresponding gene expression data has identified enrichment of (i) the immune response and AP-1 signaling pathways, and (ii) IFNG, TGFB1, PKC, Cytokine, and MAPK14 hubs.

Conclusion: While subtyping is often performed with genome-wide molecular data, we have shown that it can also be applied to categorizing histology sections. Accordingly, we have identified a subtype that is a predictor of the outcome as a result of a therapeutic regime. Computed representation has become publicly available through our Web site.

Figure 1

A pinhole view of GBM tumor section indicates a rich spatial composition in terms of nuclear size, cellularity, and presence of lymphocytes.

Figure 2

Steps in delineating each nucleus from an H&E stained tissue sections.

Figure 3

Steps in delineation of nuclei. (A) Reference image for color normalization, (B) Original H&E image, (C) normalized image, (D-E) color decomposition for each stain, (F) thresholding, and (G) refinement and validation.

Figure 4

Computational pipeline consists of four modules: downloads images from the NIH repository. Each image is partitioned into strips of (1k-by-number of columns), stored in the OMEIS image server. Each strip is partitioned into blocks of 1k-by-1k pixels, where each block is submitted to one of the two clusters at Berkeley Lab. Computed representations are then imported into a PostgreSQL database.

Figure 5

Nuclear segmentation and region-based tessellation for preferred subtypes of Figure 6E: (A) high cellularity, (B) low cellularity, (C) medium cellularity, (D) high cellularity with pleomorphism, and (E) extreme high cellularity.

Figure 6

Steps in identifying subtypes from morphological descriptors of a tissue section. (A) Each patient may have multiple tissue sections, which are accessible along with the computed features and coded clinical information through BioSig in (B). (C) Each feature, from each tissue, is represented as a density distribution that is normalized in (D). (E) Subtyping identifies 5 classes through consensus voting. (F) Following the Kaplan Meier test, only one subtype proved to have a significant p-value between pair-wise survival curves.

Figure 7

Subnetwork enrichment analysis has revealed 6 hubs with p-value < 0.05: IFNG, TGFB1, MAPK14, Cytokine, PKC, and ILB1. Union of these subnetworks and interactions indicates interactions between of these hubs.