Three-Dimensional (3D) Image Analysis of Benign vs Malignant Follicular Lymphoid Proliferations (FLPs)
AR Sohani, V Brodsky, AM Ackerman, EP Hochberg, JR Gilbertson, Y Yagi. Massachusetts General Hospital, Boston; Mass General Hospital Cancer Center, Boston
Background: Morphologic distinction between florid reactive follicular hyperplasia (RFH) and follicular lymphoma (FL) may be challenging. We studied whether 3D histologic reconstruction of scanned whole-slide serial sections could aid in this distinction by highlighting aspects of nodal architecture that are difficult to visualize on conventional 2-dimensional (2D) sections.
Design: A case each of florid RFH involving tonsil and low-grade FL involving a lymph node were selected for 3D analysis. 100 5m-thick H&E-stained serial sections were made from a formalin-fixed paraffin-embedded block of each. Serial sections were scanned with 0.33m/pixel resolution using Mirax Scan device (3DHISTECH Ltd, Carl Zeiss Microimaging GmbH); 3D reconstruction was performed using 3DHistech's Mirax Viewer software. Images of serial sections were aligned to produce a 3D stack. Limited computational power forced us to decrease the 3D model's resolution. However the model could be sectioned in various planes, freely rotated and moved, exposing 3D relationships. Spacing between scanned sections was varied dynamically during the analysis to emphasize/explore specific features.
Results: Morphologic features that were significantly enhanced upon 3D reconstruction and analysis over conventional 2D examination included: back-to-back follicles in FL vs RFH; higher volume of cortex relative to paracortex in FL vs RFH; patency of vascular sinuses in RFH vs FL; and intact mantle zones with asymmetric polar thickening in RFH vs FL. Polarization and cellular content of individual follicles (eg distinct dark/light zones, # centroblasts vs centrocytes, # mitoses and tingible-body macrophages [TBMs]) were only moderately enhanced by this analysis. The relatively low resolution of the 3D model precluded extensive analysis of cellular interactions (eg lymphocyte-dendritic cell interactions) within follicles.
Conclusions: Certain low-power morphologic features that help distinguish benign from malignant FLPs are significantly enhanced by 3D image analysis. Such analysis may be cumbersome for routine diagnostic use in straightforward cases of RFH and low-grade FL, but may help distinguish RFH from grade 3 FL which share many high-power morphologic (large # centroblasts; high # mitoses and TBMs) and immunohistochemical (high Ki67, bcl2 negativity) features within follicles. In the future, computational power will increase to allow higher resolution 3D analysis.
Tuesday, March 10, 2009 9:30 AM
Poster Session III # 146, Tuesday Morning