Joerg Bredno

Using contextual information to classify nuclei in histology images

Nucleus classification is a central task in digital pathology. Given a tissue image, our goal is to classify detected nuclei into different types, for example nuclei of tumor cells, stroma cells, or immune cells. State-of-the-art methods achieve this by extracting different types of features such as morphology, image intensities, and texture features in the nucleus regions. Such features are input to training and classification, e.g. using a support vector machine. In this paper, we introduce additional contextual information obtained from neighboring nuclei or texture in the surrounding tissue regions to improve nucleus classification. Three different methods are presented. These methods use conditional random fields (CRF), texture features computed in image patches centered at each nucleus, and a novel method based on the bag-of-word (BoW) model. The methods are evaluated on images of tumor-burdened tissue from H&E-stained and Ki-67-stained breast samples. The experimental results show that contextual information systematically improves classification accuracy. The proposed BoW-based method performs better than the CRF-based method, and requires less computation than the texture-feature-based method.

Abstract B47: Unraveling tumor metabolism with in silico IHC multiplexing supported by automated imaging analysis

The aim of our study is to understand and characterize tumor biology and its metabolism by evaluating some of the most relevant hallmarks of cancer like hypoxia, apoptosis, tumor proliferation, angiogenesis and immune status. We selected from our tissue bank FFPE blocks from 3 main tumor indications (CRC, GC, RCC) analyzed with 9 different markers: CA9, CC3, Ki67, CD34, aSMA, Podoplanin, CD3, CD8 and FOXP3 in consecutive sections. We applied 2 triplex, one duplex and one single IHC assays. After quality check made by board certified pathologist, all slides were scanned with the iScan-HT scanner (VENTANA) at 20x magnification and the whole slide sections were analyzed with help of automated imaging algorithms developed in house. Tumors harbor different metabolic states. Our cutting-edge approach enables comprehensive understanding of the interactions among different tumor components with respect to tumor metabolism and immune infiltration. Our automated image analysis tool is able to reconstruct and understand the architectural patterns of different tumor metabolic states and to show distribution of hypoxia (CA9), proliferation (Ki67), cell death (CC3), angiogenesis (CD34/aSMA/Podoplanin) and immune cells (CD8/CD3, FOXP3). Deep understanding of tumor biology and metabolism is crucial for guiding drug development and overcoming challenges in clinical translation. Citation Format: Suzana Vega Harring, Marta Canamero, Konstanty Korski, Georges Marchal, Oliver Grimm, Claudia Ferreira, Joerg Bredno, Christophe Chefd`hotel, Fabien Gaire. Unraveling tumor metabolism with in silico IHC multiplexing supported by automated imaging analysis. [abstract]. In: Proceedings of the AACR Special Conference: Metabolism and Cancer; Jun 7-10, 2015; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Res 2016;14(1_Suppl):Abstract nr B47.

Abstract 5117: An automated 5-plex fluorescent immunohistochemistry enabled characterization of PD-L1 expression and tumor infiltrating immune cells in lung and bladder cancer specimens

Cancers may escape immune surveillance and eradication through the expression of programmed death-ligand 1 (PD-L1) on tumor cells and in the tumor microenvironment. PD-L1 expression has been reported in various cell populations within the tumor, and its expression associated with prognosis for various tumors. Further, clinical studies have shown that this pathway is an important target for immunotherapy and PD-L1 expression on tumor cells and in the tumor microenvironment has been associated with enhanced response. Understanding PD-L19s complex biological function not only on tumor cells but also within the tumor microenvironment requires simultaneous interrogation of multiple biomarkers, ranging from cancer immunology checkpoint markers, tumor infiltrating immune cell markers, and tumor specific markers, etc. Multiplex immunohistochemistry (IHC) allows simultaneous detection of multiple markers to explore the potential cellular composition of immune/stromal/cancer cells in tumor microenvironment. Development of a multiplex IHC assay remains challenging due to antibody species similarity and cross reactivity, stability of fluorophores through multiple rounds of processing, balancing high and low signals and measurement of weakly expressed markers. We present here the development of a fully automated multiplex IHC assay (PD-L1, CD3, CD8, CD68 and FoxP3) using rabbit primary antibodies with a heat deactivation process between each antigen staining cycles on the BenchMark ULTRA automated slide stainer. As part of the technology validation, we compared the 5-plex IHC to the respective single-plex chromogenic IHC assays. Using the automated 5-plex fluorescent IHC assay, we tested a cohort of non-small cell lung (NSCLC) and bladder cancer tissue specimens and characterized PD-L1 and immune marker expression in both tumor and infiltrate immune cells. To provide an objective and reliable readout of the assay, image analysis tools are being developed for automated identification and quantification of the labelled biomarkers and their co-expression on a cell-by-cell basis. This automated multiplex PD-L1 5-Plex IHC assay could be utilized as a tool for further characterizing tumors and its microenvironment and gain a better understanding of which patients may benefit from immune-therapies. Citation Format: Wenjun Zhang, Antony Hubbard, Adriana Racolta, Nick Cummins, Mehrnoush Khojasteh, Liping Zhang, Karl Garsha, Joerg Bredno, Dustin Harshman, Srabani Bhaumik, Tobin Jones, Marcin Kowanetz, Sanjeev Mariathasan, Ian McCaffery, Dustin Smith, J Andrew Williams, Lidija Pestic-Dragovich, Larry Morrison, Lei Tang. An automated 5-plex fluorescent immunohistochemistry enabled characterization of PD-L1 expression and tumor infiltrating immune cells in lung and bladder cancer specimens. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 5117.

Adaptive whole slide tissue segmentation to handle inter-slide tissue variability

Automatic whole slide (WS) tissue image segmentation is an important problem in digital pathology. A conventional classification-based method (referred to as CCb method) to tackle this problem is to train a classifier on a pre-built training database (pre-built DB) obtained from a set of training WS images, and use it to classify all image pixels or image patches (test samples) in the test WS image into different tissue types. This method suffers from a major challenge in WS image analysis: the strong inter-slide tissue variability (ISTV), i.e., the variability of tissue appearance from slide to slide. Due to this ISTV, the test samples are usually very different from the training data, which is the source of misclassification. To address the ISTV, we propose a novel method, called slide-adapted classification (SAC), to extend the CCb method. We assume that in the test WS image, besides regions with high variation from the pre-built DB, there are regions with lower variation from this DB. Hence, the SAC method performs a two-stage classification: first classifies all test samples in a WS image (as done in the CCb method) and compute their classification confidence scores. Next, the samples classified with high confidence scores (samples being reliably classified due to their low variation from the pre-built DB) are combined with the pre-built DB to generate an adaptive training DB to reclassify the low confidence samples. The method is motivated by the large size of the test WS image (a large number of high confidence samples are obtained), and the lower variability between the low and high confidence samples (both belonging to the same WS image) compared to the ISTV. Using the proposed SAC method to segment a large dataset of 24 WS images, we improve the accuracy over the CCb method.