Adaptive imaging for pioneering multiplexed spatial profiling of tumors and tissues for cancer research and diagnosis

Rarecyte, Inc., Seattle WA

Investigators

Abstract

SUMMARY Histopathology plays a central role in the diagnosis and staging of cancer by ensuring that individual treatment plans correctly account for tissue of origin, tumor cell state, and immune status. Despite its importance, histopathology has changed little over the past fifty years: in both diagnostic and research settings, manual examination of hematoxylin and eosin (H&E) stained slides remains the norm with additional information provided in some cases by immunohistochemistry (IHC). However, two breakthroughs are now impacting image-based analysis of tissues and tumors: (i) methods that quantify 10-60 protein markers in tissue sections at subcellular resolution and (ii) approaches to machine learning/artificial intelligence (ML/AI) able to interpret digital images at scale. This proposal focuses on combining these advances in a fundamentally new approach to acquiring highly multiplexed tissue images from the formaldehyde fixed paraffin embedded (FFPE) specimens that are the mainstay of all pathology workflows and then computing spatial profiles. Our proposed Orion2 approach will dramatically extend an OrionTM method that we recently commercialized; Orion is able to collect 18 to 20 plex immunofluorescence data in a single round of imaging followed by diagnostic grade H&E imaging on the same section (cycling allows for higher plex analysis). Orion2 will implement an “attention-based” approach that focuses image acquisition on the most important or informative regions of a specimen. This will be accomplished by first scanning specimens at relatively low magnification and then using automated algorithms to prioritize specific tissue features for high-resolution 2D and 3D image acquisition. This type of adaptive imaging recapitulates the way in which histopathologists examine complex tissue specimens at low and high power while adding molecular detail and the automation possible with digital imaging. The Orion2 instrument will meet FDA requirements for whole-slide imaging (WSI) in diagnosis and will achieve the spatial power needed to reliably identify recurrent cell populations and spatial features of tumors associated with disease progression or drug response. Aim 1 focuses on the development of an instrument with automated control over transmitted and fluorescence illumination to allow speed of data acquisition and increased image resolution to be balanced against each other to varying degree. Aim 2 involves the development of classical and machine learning algorithms able to use low magnification scans to prioritize specific tissue features for in-depth analysis. Aim 3 links these two developments in an instrument that uses closed-loop control to collect optimized mid-plex multi- resolution images of solid tumors with an initial focus on common cancer types (breast, lung, and colorectal cancers) and mouse models of these diseases. Drawing on our experience with assay development, optics, and in vitro diagnostics, Orion2 will set a standard for image quality, ease of use, repeatability, and throughput not achieved by any existing approach and will unlock new markets in academe, industry and medicine.