DLBCL-Morph: Morphological features computed using deep learning for an annotated digital DLBCL image set

Abstract

Diffuse Large B-Cell Lymphoma (DLBCL) is the most common non-Hodgkin lymphoma. Though histologically DLBCL shows varying morphologies, no morphologic features have been consistently demonstrated to correlate with prognosis. We present a morphologic analysis of histology sections from 209 DLBCL cases with associated clinical and cytogenetic data. Duplicate tissue core sections were arranged in tissue microarrays (TMAs), and replicate sections were stained with H&E and immunohistochemical stains for CD10, BCL6, MUM1, BCL2, and MYC. The TMAs are accompanied by pathologist-annotated regions-of-interest (ROIs) that identify areas of tissue representative of DLBCL. We used a deep learning model to segment all tumor nuclei in the ROIs, and computed several geometric features for each segmented nucleus. We fit a Cox proportional hazards model to demonstrate the utility of these geometric features in predicting survival outcome, and found that it achieved a C-index (95% CI) of 0.635 (0.574,0.691). Our finding suggests that geometric features computed from tumor nuclei are of prognostic importance, and should be validated in prospective studies.

Fig. 1

Data pipeline for a single core from an H&E stained tissue microarray (TMA). In (a) the red rectangle is the pathologist-annotated ROI. In (c) red corresponds to cell nuclei classified as “neoplastic” by HoVer-Net. Green corresponds to “inflammatory” and orange corresponds to “non-neoplastic epithelial”.

Fig. 2

Tissue microarrays (TMAs) with region-of-interest (ROI) annotations. (a) H&E stained TMA. The red rectangles denote ROIs annotated by a human expert. Some missing or unrepresentative cores have no ROIs. (b) A single core from the TMA in (a) with ROI that ignores unrepresentative areas of the core. (c) BCL6 stained TMA, containing cores from the same patients as (a). (d) A single annotated core from the TMA in (c). Cells stained orange show greater BCL6 expression.

Fig. 3

Rectangle and ellipse fitted to a single segmented tumor nucleus. (a) a binary segmentation image for a tumor cell nucleus. For visual clarity, the image is zero-padded by 5 pixels on each side. (b) rotated rectangle fit to the nucleus. Our dataset provides the rectangle’s center coordinates, width, height and rotation angle. (c) rotated ellipse fit to the nucleus. Our dataset provides the ellipse’s center coordinates, perimeter, area, and major and minor axis lengths.

Fig. 4

The directory structure of DLBCL-Morph.