[1590] How Digital Image Analysis Applied to Immunohistochemistry Helps To Assess Biomarker Coexpression

Xavier Moles Lopez, Paul Barbot, Nicky D'Haene, Calliope Maris, Sandrine Rorive, Isabelle Salmon, Christine Decaestecker. Université Libre de Bruxelles (ULB), Brussels, Belgium; Center for Microscopy and Molecular Imaging, ULB, Gosselies, Belgium; Erasme Hospital ULB, Brussels, Belgium

Background: Immunohistochemistry (IHC) has allowed identification of various tissue biomarkers with diagnostic, prognostic or theranostic values. However, pathologies such as cancers combine complex interactions of different pathways, which make the study of colocalization of multiple markers of great importance (e.g. to identify cell types expressing different biomarkers). However, multiple IHC has technical limitations so that this technique is not suitable in clinical practice. Consequently, new methods to evaluate colocalization of multiple biomarkers are needed.
Design: We developed an original approach to combine information provided by multiple IHC stains on serial tissue sections. This integrated approach combines standardized IHC, whole slide scanning and digital image analysis. Expression patterns are colocalized using a two-step, multiscale, image registration procedure that was chosen for its capability to register extremely large images.
Results: This multiscale approach starts by registering pairs of images downsampled to a very low resolution (approximately equivalent to a 0.13X objective) and then refines the registration transform for gradually increasing resolutions. During the first step, resolutions are increased up to a resolution corresponding to a 2X objective (chosen as a compromise between the execution time and the registration precision). However, this step does not allow a valid colocalization of IHC stains at the histological structure scale. It is thus completed by a second step where the corresponding regions of interest (extracted from the two images) are finely registered with an increased resolution (corresponding to a 20X objective). Our approach is qualitatively and quantitatively validated using checkerboard visualization and pathologist supervision.
Conclusions: This method enables to characterize particular cell populations or to test multiple biomarkers and identifies with greater accuracy cellular heterogeneity in tissue samples. Our goal is to apply this methodology to characterize colocalization of proliferation marker (Ki67) and expression of growth factor receptors (IGF-1R, EGFR) in glioblastomas.
Category: Informatics

Wednesday, March 6, 2013 1:00 PM

Poster Session VI # 277, Wednesday Afternoon


Close Window