Machine and deep learning single-cell segmentation and quantification of multi-dimensional tissue images

Abstract

Increasingly, highly multiplexed in situ tissue imaging methods are used to profile protein expression at the single-cell level. However, a critical limitation is a lack of robust cell segmentation tools applicable for sections of tissues with a complex architecture and multiple cell types. Using human colorectal adenomas, we present a pipeline for cell segmentation and quantification that utilizes machine learning-based pixel classification to define cellular compartments, a novel method for extending incomplete cell membranes, quantification of antibody staining, and a deep learning-based cell shape descriptor. We envision that this method can be broadly applied to different imaging platforms and tissue types.

Differential pre-malignant programs and microenvironment chart distinct paths to malignancy in human colorectal polyps

“…Slides were washed in 1X DPBS, mounted in Prolong Gold (Invitrogen) and imaged using a Zeiss Axio Imager M2 microscope with Axiovision digital imaging system (Zeiss; Jena GmBH). Multiplexed imaging using an immune cell-based antibody panel was performed by using a multiplex iterative staining and fluorescence-inactivation protocol, as previously described (McKinley et al, 2017(McKinley et al, , 2019, and imaged on an Olympus X81 inverted microscope (20X magnification) with a motorized stage. For histological analysis, slides were processed and stained for hematoxylin and eosin and beta-catenin using standard approaches.…”

“…MxIF, single-cell segmentation and image analysis Cell segmentation was accomplished using the MANDO pipeline (McKinley et al, 2019). Briefly, random forest pixel classification on manually annotated images was used to define tissue and subcellular regions in each image.…”

Quantifying and correcting slide-to-slide variation in multiplexed immunofluorescence images

“…The data was collected from human colorectal cancer tissue samples from the Human Tumor Atlas Network (Rozenblatt-Rosen et al, 2020). The final dataset comprises over 2.2 million cells in the MxIF modality across over 2400 images on 43 different slides, with single-cell segmentation performed using an algorithm developed in-house (McKinley et al, 2019). Cell intensities for each marker were quantified as the median pixel value within the segmented cell, with tissue samples stained for 33 different marker channels.…”

“…Single-cell assays are increasingly valued for their ability to provide information about the cell microenvironment and cell population interactions in healthy and cancerous tissues (Islam et al, 2020;McKinley et al, 2019;Shrubsole et al, 2008). Multiplexed imaging methods like multiplexed immunofluorescence (MxIF) (Gerdes et al, 2013), multiplexed immunohistochemistry (IHC) (Tsujikawa et al, 2017) and CODEX (Goltsev et al, 2018) are in situ analyses of multiple marker channels over a large number of cells within a given tissue sample.…”

Use of Single-Cell -Omic Technologies to Study the Gastrointestinal Tract and Diseases, From Single Cell Identities to Patient Features

“…Using multiplex protein imaging, Ligorio et al 28 found that cancerassociated fibroblasts contribute to heterogeneity within pancreatic tumors. Algorithms to analyze multidimensional images can be segmentation-based to produce single-cell resolution data similar to mass cytometry 29,30 or pixelbased. 31 The unifying theme behind these studies revolves around the discovery of new, unexpected cell populations that associate with disease processes.…”