Quality control and capibration devices and methods for microscopes to enable single-cell resolution spatial transcriptomics

Abstract

A method for compensating for quality variation in an imager capturing microscope images or sequencing images associated with a tissue is disclosed. A plurality of images of an optical target, captured by an imager, the optical target including a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region, the plurality of images including an image of each different area of the distortion control region, are provided. One or more of the images of the different areas of the distortion control region are fed to a distortion correction process configured to perform a distortion correction based on locations of one or more of the patterned fiducial regions, and profile data is generated for the imager, the profile data comprising the distortion correction data from the distortion correction process. The quality variation profile data is then stored for use in correcting subsequent microscope images or subsequent sequencing images received from the imager. An optical target is also disclosed. The optical target includes a metal patterned substrate having a blank region, a focusing zone region comprising a plurality of focusing features, and a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region.

Description

33080 / IP-2902 / PCIP-3041 -PCTQUALITY CONTROL AND CAPIBRATION DEVICES AND METHODS FOR MICROSCOPES TO ENABLE SINGLE-CELL RESOLUTION SPATIAL TRANSCRIPTOMICSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to (1 ) U.S. Provisional Patent Application No. 63 / 705,910, filed October 10, 2024, entitled “Tissue Image Stitching and Registration for Spatial Transcriptomics,” (2) U.S. Provisional Patent Application No. 63 / 705,975, filed October 10, 2024, entitled “Substrates and Related Methods and Systems,” and (3) U.S. Provisional Patent Application No. 63 / 761 ,850, filed February 21 , 2025, entitled “Quality Control and Calibration Devices and Methods for Microscopes to Enable Single-cell Resolution Spatial Transcriptomics,” the entire disclosures of each of which is hereby expressly incorporated by reference herein.FIELD OF THE INVENTION

[0002] The present disclosure generally related to techniques for quality control and calibration in microscope images or sequencing images associated with a tissue.BACKGROUND

[0003] The background description provided herein is for the purpose of generally presenting the context of the disclosure. Work of the presently named inventors, to the extent it is described in this background section, as well as aspects of the description that may not otherwise qualify as prior art at the time of filing, are neither expressly nor impliedly admitted as prior art against the present disclosure.

[0004] In ex situ spatial transcriptomics, transcripts are obtained from tissue samples placed on a substrate after tissue permeabilization and subsequent capture using target capture sites on the surface. The captured transcripts are then assigned locations during cDNA synthesis to generate a spatial transcriptomics map. Prior to tissue permeabilization, the tissue can be stained and imaged using a microscope. The microscope image may then depict cells, nuclei, cell morphology, tissue histology, and / or other features of the tissue sample.

[0005] For ex situ spatial transcriptomics platforms, the level and distribution of a single FOV distortion from the brightfield microscope tissue images is mostly co-determined by microscope lens used and camera sensor specs, but generally undisclosed by microscope vendors to users. Therefore, its impact on feature-based image stitching accuracy or subsequent spatial heatmap alignment precision is not monitored or well characterized on microscope histology images used for spatial transcriptomics platforms. On-market solutions have yet to address this issue,33080 / IP-2902 / PCIP-3041 -PCT partially due to general difficulty to support different third-party microscopes from various vendors with different specs and settings, while partially due to lack of systematic understanding on different types of error distributions that lead to incorrect spatial information and low resolution in the platform’s assay workflow. The optical target of the present invention is the first among all spatial platforms to fully evaluate the impact of distortion and image quality from brightfield microscopes based on simple designs above, addressing technical difficulties above with easy microscope user experience that requires no pre-requisite for lay person. The present invention may also be compatible with fluorescence microscopes to improve their imaging performance and / or analysis of images captured therefrom.

[0006] Integrating tissue image stitching and registration with spatial transcriptomics has become essential for generating comprehensive spatial maps of gene expression patterns within complex tissue structures. To precisely map out the transcription information, a critical step is to align the spatial genomics information from the sequencing with the tissue staining images from the microscopy. The major challenges for accurate stitching and registration include image resolution, distortion and variability, and identifying an appropriate algorithm and approach to accurately assemble all of the sub-images together.SUMMARY

[0007] Three metrics to evaluate any spatial transcriptomics platforms are sensitivity, resolution, and multiplexity. The definition of resolution limit is the minimal distance of where the transcript signal is detected compared to where the transcript signal is in a tissue sample. Fundamentally, this limit is dictated by several factors, including optical resolution of spatial barcodes (identifiable grid size), sensitivity, and in many ex situ spatial platforms, the severity of mRNA diffusion in the workflow during and after total deconstruction of cellular environment (in general, permeabilization and tissue digestion) which ultimately affects the accuracy of transcript localization during capturing. In practice, due to limitations of assay performance on sensitivity and mRNA diffusion, users generally take stained histology images under microscopes before proceeding to tissue deconstruction and use it as the reference map to overlap and align with the transcript heatmap that is reconstructed from the full assay workflow. This enables more accurate spatial location of transcripts at the tissue organization and histology structure level. Therefore, the level of precision on the alignment between the reconstructed transcript heatmap and the microscope staining tissue image is valuable in determining the real spatial resolution of the ex situ assay.33080 / IP-2902 / PCIP-3041 -PCT

[0008] There are four main aspects to be considered for accurately and precisely overlapping and aligning these two images, i.e., the reconstructed transcript heatmap and the microscope staining tissue image. The main aspects include: a) (linear and mostly nonlinear) distortion (and its correction) of each field-of-view (FOV) microscope image, b) sufficient overlap between each of the neighboring FOV microscope images for image stitching to generate accurate tissue-size large microscope images, c) impact mainly from lateral diffusion, detection sensitivity and barcode accuracy on the precision of re-constructed transcript heatmaps, and d) image registration through specific anchoring points that provide “ground truth” of the image features’ physical location, which are also generally known as “fiducials”.

[0009] For sequencing systems and flow cells, fiducial registration has been discussed and used to address feasibility and performance. Fiducial registration is used for distortion correction for the transcript heatmaps that each FOV is called a “tile” in a sequencer’s primary analysis pipeline. It is widely acknowledged, for example, that the precision of a reconstructed transcript heatmap is largely limited by detectability (sensitivity) and diffusion. Most microscope vendors address stitching by setting image overlap portion no less than 10% and can be tuned by users. Single FOV images, however, remain relatively poorly characterized for “quality control” purposes in the common practice of spatial transcriptomics applications, especially in ex situ platforms. This is mostly due to 1 ) the level of complexity and difficulty to support various types and brands of microscopes including different objectives, camera sensors, and imaging methods; and 2) the high level of operational freedom in the hands of microscope users when taking histology tissue images during the assay workflow. More common strategies adopted by vendors who develop spatial transcriptomics platforms in the field are 1) listing recommended imaging vendors and respective systems as well as specifying general imaging configuration recommendations, and / or 2) developing and producing an in-house imaging system that has been evaluated using the vendor’s specific assay workflow and allows minimal user interference or input for imaging conditions.

[0010] Yet still variations in microscopes results in distortion in microscope images that are difficult to sufficiently remove. Generally speaking, distortions on microscope images are mostly attributed to objective lens and its alignment. Microscope vendors in general do not list distortion as a common objective specification externally, nor do microscope vendors make distortion information available to their users. The general knowledge by optics experts in the field is that vendors control distortion level within ~1% per FOV size so it does not get detected in visual check by users. Unfortunately, this threshold is not always serving as sufficient information to33080 / IP-2902 / PCIP-3041 -PCT support spatial transcriptomics platforms, especially ex situ platforms, to reach single-cell resolution, which is typically considered —5-10 micron. For example, a 20X magnification objective with NA 0.4-0.8 is typically required on brightfield microscopes to achieve single-cell level feature recognition on the stained tissue samples. Further, due to recent development of cameras with large number of pixels and wide-views, the field of view (FOV) size for capturing microscope images could range up to 1 -2 millimeters. Such large FOVs could easily generate -10-20 microns range of non-linear distortion towards the edge of the FOV. Such distortion values pose a more serious concern for image stitching on FOV tiles, as stitching in general utilizes common features from shared borders of neighboring FOV tiles with an overlap of 10- 30% area on the edge where distortion is the worst on each FOV.

[0011] To combat these challenges, the present disclosure demonstrates an optical target (OT) that can be used as a universal quality control (QC) and calibration tool that can be directly applied on different microscopes from various vendors to measure and correct the lens distortion, address image focus issue, and detect single FOV tile shading uniformity to assist with image stitching and registration for large areas. More details about image stitching and registration may be described in U.S. Provisional Patent Application No. 63 / 705,910, titled “Tissue Image Stitching and Registration for Spatial Transcriptomics,” filed October 10, 2024, which is hereby incorporated by reference, in its entirety.

[0012] In one aspect, a method for compensating for quality variation in an imager capturing microscope images or sequencing images associated with a tissue is disclosed. The method includes: receiving a plurality of images of an optical target, captured by an imager, the optical target comprising a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region, wherein the plurality of images comprises an image of each different area of the distortion control region; providing one or more of the images of the different areas of the distortion control region to a distortion correction process configured to perform a distortion correction based on locations of one or more of the patterned fiducial regions; generating profile data for the imager, the profile data comprising the distortion correction data from the distortion correction process; and storing the quality variation profile data for use in correcting subsequent microscope images or subsequent sequencing images received from the imager.

[0013] In another aspect, an optical target is disclosed. The optical target includes: a metal patterned substrate having a blank region, a focusing zone region comprising a plurality of focusing features, and a distortion control region formed of a plurality of patterned fiducial33080 / IP-2902 / PCIP-3041 -PCT regions each defining a different area of the distortion control region. More details about the metal patterned substrate and patterned fiducial regions may be described in (1 ) U.S. Provisional Patent Application No. 63 / 705,975, entitled “Substrates and Related Methods and Systems,” filed October 10, 2024 and (2) U.S. Provisional Patent Application No. 63 / 765,305, entitled “Substrates and Related Methods and Systems,” filed February 28, 2025, each of which is hereby incorporated by reference in its entirety.BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The figures described below depict various aspects of the system and methods disclosed herein. It should be understood that each figure depicts an embodiment of a particular aspect of the disclosed system and methods, and that each of the figures is intended to accord with a possible embodiment thereof.

[0015] There are shown in the drawings arrangements which are presently discussed, it being understood, however, that the present embodiments are not limited to the precise arrangements and instrumentalities shown, wherein:

[0016] FIG. 1 depicts a block diagram of a workflow using an optical target to correct microscope-based field of view (FOV) errors, according to some aspects.

[0017] FIGs. 2A-2F illustrate various views and schematics of an optical target slide and components thereof, according to some aspects.

[0018] FIG. 3 depicts an example process for transforming the coordinates in a microscope image to a ground truth coordinate system using fiducials at known locations, according to some aspects.

[0019] FIG. 4A depicts an example field of view of a microscope image registered to a ground truth coordinate system using fiducials, according to some aspects.

[0020] FIG. 4B depicts another example field of view of a microscope image registered to a ground truth coordinate system using uniquely patterned fiducials, according to some aspects.

[0021] FIG. 5 depicts an example fiducial design with fiducials having unique features for identifying the fiducials in a microscope image or spatial transcript map, according to some aspects.

[0022] FIG. 6 depicts another example fiducial design with fiducials having unique features for identifying the fiducials in a microscope image or spatial transcript map, according to some aspects.33080 / IP-2902 / PCIP-3041 -PCT

[0023] FIG. 7 depicts an example software algorithm for registering a microscope image to a ground truth coordinate system and performing an error assessment, according to some aspects.

[0024] FIG. 8 depicts another example software algorithm for registering a microscope image to a ground truth coordinate system and performing an error assessment, according to some aspects.

[0025] FIG. 9 depicts example hybrid substrates having nanoimprint lithography (NIL) and metal fiducials, according to some aspects.

[0026] FIG. 10 depicts a detailed view of one of the hybrid substrates shown in FIG. 9 having NIL fiducials and metal fiducials, according to some aspects.

[0027] FIG. 1 1 depicts examples of distortion correction results, according to some aspects.

[0028] FIGs. 12A-12C depict additional examples of distortion correct results for different patterns of fiducials, according to some aspects.

[0029] FIG. 13 depicts example distortion profiles for small and medium fields of view, according to some aspects.

[0030] FIG. 14 depicts an example plot of registration score vs. pixel error size, according to some aspects.

[0031] FIG.15 depicts an example distortion estimation workflow, according to some aspects.

[0032] FIG. 16 depicts a network diagram of an exemplary network environment which may be used to perform the methods disclosed herein, according to some aspects.DETAILED DESCRIPTION

[0033] Fiducials can be placed on a substrate or flow cell as reference points to locate tissue regions and register clusters in sequencing. To effectively use the fiducials as reference points, the fiducials must be visible under conventional bright field microscope for tissue image registration, and fluorescence-visible on sequencers for cluster registration. However, some fiducials can be difficult to detect under a microscope when a tissue is loaded on the substrate, because the fiducials may have poor contrast in bright field, especially when they are covered by a tissue. In this instance, light scattering and refraction are interfered with so that the fiducials are not visible.33080 / IP-2902 / PCIP-3041 -PCT

[0034] To combat these challenges, an innovative model system integrates metal -patterned fiducials into the image alignment workflow. This approach can be used to better characterize the distortion error and precisely register and stitch the images to the reference coordinates, which that contains the theoretical XY location for each sub-image. More details about registering a microscope image to reference coordinates, e.g., a ground truth coordinate system, may be described in U.S. Provisional Patent Application No. 63 / 735,535, entitled “Spatial Image Registration to Ground Truth,” filed December 18, 2024, which is hereby incorporated by reference in its entirety.

[0035] Metal-patterned fiducials can be fabricated onto a substrate using photolithography techniques. These techniques ensure the pattern resolution down to 1 -2 pm in feature size and 10 nm positioning accuracy. The fiducials can be placed at known physical locations within the substrate.

[0036] The metal-patterned fiducials can include a photoresist layer on top of a metal layer. Because of its low optical transmission, the metal layer shows dramatically enhanced contrast to the tissue background, providing great visibility under a bright field microscope. The dark patterns from the metal layer can be used as coordinates to align and stich images of the tissue sample, while the photoresist layer can emit fluorescence under excitation of laser, which results in bright patterns to register clusters on sequencers. With this combination, the fiducials can be sufficiently visible under an optical microscope and a sequencer. One example combination of a metal layer and a photoresist layer is chromium with epoxy resin-based photo patternable material (PPM).

[0037] The metal-patterned fiducials can also have a geometry for identifying the fiducials. For example, the fiducials may have a cross, line, or bullseye ring geometry. The line pattern may be used to achieve a higher density of fiducials on the substrate. The bullseye pattern enables the substrate to be rotated and for the orientation of the bullseye to not be substantially affected and / or for the bullseye to be detected if the substrate is rotated

[0038] In some implementations, the pattern of fiducials are in an asymmetrical pattern, a random pattern, and / or a scattered pattern. In some implementations, the fiducials are irregularly spaced and / or may have different sizes. In some implementations, the fiducials may be sporadically spaced and / or may not arranged with consistent alignment or along a grid. In some implementations, the fiducials are arranged in a non-uniform manner.33080 / IP-2902 / PCIP-3041 -PCT

[0039] Also in some implementations, the fiducials may include unique features so that they are uniquely identifiable. For example, each fiducial may include eight rings with one ring in the center and seven rings surrounding the center ring in an outer circular pattern. Each of the seven rings may or may not have metal or a nanowell in the center which is used as an encoding scheme to uniquely identify the fiducial. This allows for 2A7 or 128 unique vectors. In this manner, the system can uniquely identify the fiducial according to the encoding scheme and determine its associated known physical location. In another example, the system may uniquely identify a fiducial based on its distance from other nearby fiducials. For example, one fiducial may be 10 pm and 20 pm from the two nearest fiducials while another fiducial may be 15 pm and 22 pm from the two nearest fiducials. Accordingly, the system can uniquely identify the fiducial according to the distances and can determine the known physical location associated with the particular fiducial. In yet another example, the substrate may include a random combination of fiducial geometries such as crosses and rings and the random pattern may be used to uniquely identify a fiducial.

[0040] Then when a tissue sample is placed on the substrate, the tissue sample may be imaged with the metal-patterned fiducials to generate a coordinate system for the image of the tissue sample. For example, the system may capture images of the tissue sample at several fields of view (FOV), where each of the FOVs combine to depict a complete view of the tissue sample.

[0041] The present techniques demonstrate an optical target (OT) that can be used as a universal quality control (QC) and calibration tool that can be directly applied on different microscopes from various vendors to measure and correct the lens distortion, address image focus issue, and detect single field of view (FOV) tile shading uniformity to assist with image stitching and registration for large areas (see FIGs.1 and 2). The block diagram 100 in FIG.1 is a workflow using an optical target to correct microscope-based FOV errors by achieving improved fiducial-based and / or feature-based image registration, in particular as may be used in a spatial transcriptomics (ST) workflow capable of generating an overlay of ST data and registered microscope image, such as a brightfield image. The optical target may also be used with registered microscope images from fluorescence imagers to improve their imaging performance and / or analysis of images captured therefrom. More discussion on how this may be achieved is below.

[0042] At block 102, sequencing tile images 104 are provided. Fiducial detection, indicated by arrow 106, is performed based on fiducial template 108, such that per-tile fiducial locations may33080 / IP-2902 / PCIP-3041 -PCT be identified and decoded, resulting in image 110 which illustrates both the ST data and the fiducial locations, indicated by dots overlayed on the ST data.

[0043] At block 1 12, a microscope image 114 (shown at 20x resolution) is provided. An FOV distortion correction profile is determined for the image 114 based on results from a plurality of test images captured by the microscope of an optical target. Image 114 may be, for example, a brightfield image. Fiducial detection, indicated by arrow 1 16, is performed based on fiducial template 108, resulting in image 118 which illustrates both the microscope data and the fiducial locations, indicated by dots overlayed on the microscope data. FOV focus and image quality evaluation may also be performed on the image 114. An affine transformation matrix, indicated by arrow 120, is applied to the microscope image 118 based on the ST data. The affine transformation matrix may include translation, rotation, and / or scaling information between the locations of the fiducials in the image 118 and their known physical locations. FOV shading uniformity correction may also be performed on the image 118. More detail about the affine transformation matrix is illustrated in FIG. 3. Although this disclosure generally refers to an affine transformation, those skilled in the art will understand that the transformation may comprise a projective correction or may comprise only the sub-components of an affine correction, such as a scale correction, a skew correction, or a translation correction.

[0044] At block 122, the ST data and the microscope image are overlayed to produce overlayed image 124.

[0045] FIG. 2A illustrates a top view 200A of an optical target slide 200 in accordance with various examples herein. In the illustrated example, the physical optical target slide 200 has a width W of 25 mm and a length L of 75 mm. The physical optical target slide includes of an active area 200C in the center and an even layer of metal-coating, such as chromium-coating, non-active area 206 around the active area 200C. This design is physically enabled by metal patterning, which is commonly used in micro-fabrication and semi-conductor industry.

[0046] FIG. 2B illustrates a side view 200B of the optical target slide 200 combined with mounting media 202 and coverslip 204. In the illustrated example, the optical target has the shape and dimension of a standard microscope slide, so it can be used in the same way as a microscope slide with mounting media 202 (permanent or removable) and standard No.1.5 coverslip 204 (0.17 mm glass) during imaging under a microscope.

[0047] FIG. 2C illustrates the active area 200C in more detail, rotated 90-degrees clockwise from the orientation shown in FIG. 2A. The active area 200C contains 3 neighboring regions of33080 / IP-2902 / PCIP-3041 -PCT metal-patterned, such as chromium-patterned, tiles next to each other: a distortion correction (DC) tile 210 to measure and correct distortion in the center, a blank (transparent, e.g., made of optical glass, quartz, or another suitable material) tile 212 to evaluate illumination / FOV shading uniformity on one side, and a pinhole tile 214 (consisting of a hexagonal pinhole grid with 1 m diameter and 3 pm pitch) as a focusing region used to calculate its focus and level of tilt on the other side. When in use of the optical target, the user takes a through-focus Z-stack image set of the DC tile 210, and an FOV image of the blank tile 212 along each focal plane of the DC tile 210 under the target microscope in brightfield mode to evaluate focus, distortion, and shading uniformity of that same microscope. The minimal number of FOV images needed to evaluate shading uniformity may be one FOV image of the blank tile 212 and one in-focus image of the DC tile 210. The minimal number of FOV images needed to evaluate focus is a through-focus Z- stack images set of the DC tile 210 or the pinhole tile 214. The option of both a through-focus Z- stack image set of the DC tile 210 and a through-focus Z-stack image set of the pinhole tile 214 is not necessary but is recommended to improve the robustness of focus evaluation. The active area 200C is suitable for 450-2200 pm FOV sizes and 0.15-0.5 pm / pixel images. While described for use in example brightfield imager applications, the optical target slides herein may be configured to provide quality control analysis of other imagers, including fluorescence imagers. For example, the optical target slide 200C may be used to perform quality control of a fluorescence microscope by coating the optical target slide 200C with a fluorescent dye before images or the active area 200C are captured, i.e., resulting in fluorescent images of the active area 200C. Suitable fluorescent dyes may include, e.g., cyanine dye 3 (Cy3), cyanine dye 5 (Cy5), or other similar fluorophores. Additional processing is not needed to compensate for the fact that the image of the sample is from a fluorescence microscope rather than a brightfield microscope. That is, the present techniques may be implemented through optical target slides that reflect incident light (such as targets with metal fiducials) and optical target slides that emit light (such as targets with fluorescent dyes).

[0048] FIG. 2D illustrates the DC tile 210 with relevant regions outlined and labeled. The universal measurement and correction of distortion on different microscopes with varied FOV sizes and pixel sizes is achieved through the in-focus imaging of the DC tile 210. In the illustrated example, the DC tile 210 includes of a set of three overlapping regions 220, 230, and 240 to address different FOV sizes from various microscope vendors, with the largest region 240 addressing FOV up to 2.2 mm in length on x / y. Small region 220 and medium region 2 address smaller FOV sizes. Each of the regions includes four same-size dual-ring fiducials 21833080 / IP-2902 / PCIP-3041 -PCT in a square shape pattern and a hexagonal grid of pinhole arrays with the same spec as the pinhole tile 214 in the background.

[0049] FIG. 2E illustrates a close-up of one of the dual-ring fiducials 218 from DC tile 210, including the hexagonal pinhole grid. The hexagonal pinhole grid, without the accompanying dual-ring structure, is also included in the pinhole tile 214. The four dual-ring fiducials 218 from each region 220, 230, and 240 share the same inner- and outer- diameters, but vary differently among the three overlapping regions, to ensure that for any size of microscope FOV (square or rectangle shape) between 450 pm (in short axis) to 2.2 mm (in long axis) with image pixel size between 0.15 pm to 0.5 pm on a 20X or higher-magnification on-market objective, there is at least one set of four dual-ring fiducials to be registered for their locations, and subsequently used for linear translocation (including translation, shear, magnification, rotation) and correction. The pinhole arrays in the background are used to measure the actual distortion by calculating differences between their expected location (x0,y0) based on design template and their actual location (x,y) on that in-focus microscope FOV image (shown as bright spots). By recording the coordinate differences for sampled pinholes across the entire FOV, a heatmap of non-linear distortion from this same FOV can be generated, and by applying the reverse translocation using its approximate polynomial function and mapping table, distortion could be corrected (in average below 0.5 pm) on that FOV (shown in FIG. 11 ).

[0050] FIG. 2F illustrates a schematic 200F combining the details of FIGs. 2C and 2D. As shown in FIG. 2F, the fiducials 218 and the regions 220, 230, and 240 are centered on the DC tile 210. The schematic 200F is not necessarily drawn to scale.

[0051] FIG. 3 illustrates an example process for transforming the coordinates in a microscope image, such as microscope image 1 14, to physical locations using the metal-patterned fiducials, such as fiducials 218. For each FOV, the system may map locations within the image to physical locations using the metal-patterned fiducials. More specifically, the system may generate a distortion coefficient and affine transformation matrix which includes translation, rotation, and / or scaling information between the locations of the metal-patterned fiducials in the image and their known physical locations. To transform the locations of the fiducials, a ground truth calculation engine generates affine transformation matrices for the FOVs in the microscope image and the sequencing tiles in a spatial transcript map. For example, to calculate a transform and generate a first affine transformation matrix for a first FOV in the microscope image, the ground truth calculation engine establishes correspondences between a first set of observed fiducial locations in the first FOV and a first set of known locations for the fiducials. Using the33080 / IP-2902 / PCIP-3041 -PCT correspondences, the ground truth calculation engine solves for the parameters of the affine transformation, which includes translation, rotation, scaling, and shearing factors. The resulting first affine transformation matrix encapsulates the parameters, enabling the conversion of locations of fiducials within the FOV to corresponding known locations.

[0052] In some implementations, the affine transformation matrices are calculated using unconstrained least squares based on minimizing the sum of the squares of the correspondences between the observed fiducial locations and the known fiducial locations. In other implementations, the affine transformation matrices are calculated using constrained least squares (e.g., if the transformation is within a known range). In still other implementations, the affine transformation matrices are calculated using weighted least squares, where the weights are proportional to confidence levels or correlation peaks reflecting the accuracy of a corresponding fiducial detection. Then the ground truth calculation engine applies the affine transformation matrices to the FOVs in the microscope image and the sequencing tiles in the spatial transcript map. This involves manipulating the FOVs and sequencing tiles according to the parameters defined in the respective affine transformation matrices. In some implementations, the ground truth calculation engine adjusts the position, orientation, and / or scale of the FOVs and sequencing tiles to register them to the ground truth coordinate system. The transformation ensures that the fiducials and / or stitched nucleotide cluster locations (SBCs) in the spatial transcriptomics map are positioned at their known locations, facilitating accurate registration. In some implementations, the resolution of each image may be different. For example, the sequencing tiles map may have a pixel resolution of x pixels per pm, and the FOVs may have a pixel resolution of y pixels per pm. In further implementations, the ground truth coordinate system may have another different resolution (e.g., 1 pixel per pm). In some such implementations, a server device may scale the images and / or fine-tune the affine transform based on the differences. In some implementations, only sub-components of an affine transformation may be applied, such as a scale correction, a skew correction, or a translation correction. In some implementations, a projective transformation may be applied.

[0053] While the fiducials described herein may be referred to as “metal-patterned” fiducials, this is one type of fiducial material which may be used in the system. Fiducials may also include nanowell fiducials, or any other suitable fiducials using nanoimprint lithography (NIL). The fiducials may include in-tissue fiducials which are located within a portion of the imaging area that includes the tissue sample (also referred to herein as a “tissue area”). Some substrates may be referred to as “hybrid substrates” that include both NIL fiducials and metal fiducials. For33080 / IP-2902 / PCIP-3041 -PCT example, the in-tissue fiducials may include metal fiducials and the fiducials outside of the tissue area may include NIL fiducials. Further still, in some examples, NIL fiducials may contain fluorophores (such as from a sequencing sample) that emit light or fluorescent dyes, e.g., compounds that may contain one or more fluorophores.

[0054] After each FOV or sequencing tile is transformed, the FOV or sequencing tile is placed on a ground truth map by aligning the FOV or sequencing tile with the other FOVs or sequencing tiles according to the known locations of respective fiducials, as shown in FIGs. 4A and 4B.

[0055] FIG. 4A depicts an example FOV 402 of a microscope image 406 registered to a ground truth coordinate system 400. The ground truth coordinate system 400 is depicted as a grid with the fiducials 404 from the microscope image 406 at their known locations.

[0056] In some implementations, the ground truth calculation engine registers the FOV 402 to the ground truth coordinate system 400 by transforming the observed locations in the FOV 402 into known locations and stitching the FOV 402 to adjacent FOVs 402 within the microscope image 406 using detected fiducials 404 which overlap in adjacent FOVs 402. As mentioned above, the ground truth calculation engine may transform the locations in the FOV 402 by calculating a transform and generating an affine transformation matrix for the FOV 402 using correspondences between observed fiducial locations in the FOV 402 and corresponding known fiducial locations in the FOV 402. Using the correspondences, the ground truth calculation engine solves for the parameters of the affine transformation, which includes translation, rotation, scaling, and shearing factors. Then the ground truth calculation engine applies the affine transformation matrix to the FOV 402 to adjust the position, orientation, and / or scale of the FOV 402.

[0057] In some implementations, each fiducial has a known, global location within the microscope image 406. In this manner, the FOV 402 can be mapped to a particular cell (e.g., a row and column) within a grid in the ground truth coordinate system 400 based on the known, global locations of the fiducials within the FOV 402. While the fiducials are shown in FIG. 4A as having a uniform spacing and a uniform geometry (dots), this is merely one example for ease of illustration only.

[0058] The fiducials may have different spacings from each other, such that a fiducial or group of fiducials may be uniquely identified within the flow cell based on the spacing from adjacent fiducials. In this manner, the FOV 402 can be mapped to a particular cell (e.g., a row33080 / IP-2902 / PCIP-3041 -PCT and column) within a grid in the ground truth coordinate system 400 based on the spacing of the fiducials in the FOV 402. Still further, the fiducials may have different geometries from each other. Some fiducials may have a dot geometry while others have a cross, line, or bullseye ring geometry. A fiducial or group of fiducials may be uniquely identified within the flow cell based on the geometries of the fiducials. In this manner, the FOV 402 can be mapped to a particular cell (e.g., a row and column) within a grid in the ground truth coordinate system 400 based on the geometries of the fiducials in the FOV 402. In other implementations, an FOV 402 can be mapped to a particular cell (e.g., a row and column) within a grid in the ground truth coordinate system 400 based on any suitable combination of the spacing and geometries of the fiducials in the FOV 402.

[0059] For example, as shown in FIG. 4B, some fiducials include dot geometries 410 while other fiducials include cross geometries 408. The ground truth calculation engine may determine the global locations of the fiducials within an FOV based on the arrangement of geometries of the fiducials. For example, an FOV with three dot fiducials 410 above three cross fiducials 408 may be in the lower right corner of the ground truth coordinate system.

[0060] In other implementations, the fiducials have known local locations within the FOV 402. In yet other implementations, the ground truth calculation engine obtains a known spacing between the fiducials which can be used to derive local locations of the fiducials within the FOV 402. The FOV 402 may also have an identifier such as an FOV ID which can be used to obtain information regarding the spatial relationship of the FOV 402 within the microscope image 406. For example, the FOV ID may indicate whether the FOV 402 is situated in the top left cell of the ground truth coordinate system 400, directly below the top left cell, directly to the right of the top left cell, or in another specific location.

[0061] In some implementations, the ground truth calculation engine may stitch the FOV 402 to adjacent FOVs to the cell where the FOV 402 is located. For example, the FOV 402 may include at least some overlap with adjacent FOVs. As such, the ground truth calculation engine may identify common features within the adjacent FOVs and assign the common features the same or a similar physical location in the ground truth coordinate system 400. For example, the ground truth calculation engine may identify the same fiducials on the right side of one FOV and on the left side of an adjacent FOV. The ground truth calculation engine may assign the fiducials the same physical location in the ground truth coordinate system 400. More specifically, if the FOV 402 includes fiducials having known, local locations, and a first fiducial has local location (2 pm, 50 pm) which matches with or is similar to a second fiducial in an33080 / IP-2902 / PCIP-3041 -PCT adjacent FOV having global location (300 m, 320 pm) after stitching, the FOV 402 may be stitched to the adjacent FOV such that the first fiducial has global location (300 pm, 320 pm) or a similar location in the ground truth coordinate system 400. In another example, the ground truth calculation engine may stitch the FOVs based on common image features, such as cells, nuclei, etc. within the adjacent FOVs.

[0062] In any event, the ground truth calculation engine may stitch each of the FOVs 402 to each other to register the microscope image 406 to the ground truth coordinate system 400. The ground truth calculation engine may perform a similar process for registering the spatial transcript map to the ground truth coordinate system 400. For example, the ground truth calculation engine may transform the observed locations in each sequencing tile into known locations and stitch the sequencing tiles to adjacent sequencing tiles within the spatial transcript map using fiducials which overlap in adjacent sequencing tiles and / or SBCs which overlap in adjacent sequencing tiles.

[0063] In yet other implementations, the ground truth calculation engine obtains a known spacing between the fiducials which can be used to derive local locations of the fiducials within the sequencing tile. The sequencing tile may also have an identifier such as a tile ID which can be used to obtain information regarding the spatial relationship of the sequencing tile within the spatial transcript map (e.g., a swath number of a lane, an order of the tile within a swath or lane, etc.). For example, the tile ID may indicate whether the sequencing tile is the top left sequencing tile in the ground truth coordinate system 400, directly below the top left sequencing tile, directly to the right of the top left sequencing tile, or in another specific location.

[0064] The ground truth calculation engine may stitch the sequencing tile to adjacent sequencing tiles. The sequencing tile may include at least some overlap with adjacent sequencing tiles. In this manner, the ground truth calculation engine may identify common SBCs or fiducials within the adjacent sequencing tiles and assign the common SBCs or fiducials the same physical location in the ground truth coordinate system 400. For example, the ground truth calculation engine may identify the same SBCs on the right side of one sequencing tile and on the left side of an adjacent sequencing tile. The ground truth calculation engine may assign the SBCs the same or a similar physical location in the ground truth coordinate system 400.

[0065] FIGs. 5 and 6 illustrate example fiducial designs for uniquely identifying a fiducial and / or uniquely identifying a particular area {e.g., a cell) to which a group of fiducials in an FOV or sequencing tile correspond. As shown in FIG. 5, the fiducials have a known horizontal (260 pm) and vertical spacing (300 pm) from each other. Additionally, the substrate includes33080 / IP-2902 / PCIP-3041 -PCT randomly distributed fiducials in a bullseye pattern 502 having 8 rings with 1 ring in the center and seven rings surrounding the center ring in an outer circular pattern. Each of the seven rings may or may not have metal or a nanowell in the center 504a-504g which is used as an encoding scheme to uniquely identify the fiducial. This allows for 2A7 or 128 unique vectors.

[0066] The server device may identify a particular area or cell to which an FOV or sequencing tile corresponds based on the positions of the bullseye fiducials in the FOV or sequencing tile, such as microscope image 1 14 or sequencing tile 104, relative to each other. The server device may also uniquely identify a fiducial (and determine its corresponding location) based on which of the seven rings have metal in the center. In some implementations, the server device transforms the observed fiducial locations in an FOV or sequencing tile to local known locations based on the known spacing between fiducials in an FOV or sequencing tile. The server device transforms the local known fiducial locations in an FOV or sequencing tile to global known locations based on uniquely identifiable fiducials and / or uniquely identifiable groups of fiducials.

[0067] FIG. 6 illustrates another example fiducial design with a combination of bullseye fiducials and randomly distributed cross fiducials. As shown in FIG. 6, the fiducials have a known horizontal (260 pm) and vertical spacing (150 pm) from each other. For a particular FOV or sequencing tile, the server device may identify the particular area or cell to which the FOV or sequencing tile corresponds based on the positions of the cross fiducials in the FOV or sequencing tile relative to each other.

[0068] FIGs. 7 and 8 illustrate example algorithms 800, 900 for the ground truth calculation engine to register a microscope image or spatial transcript map to a ground truth coordinate system and perform an error assessment. By combining the strong fiducial contrast with micropatterned metal and the imaging processing algorithms 800, 900 depicted in FIGs. 7 and 8, submicron sub-image placement accuracy can be achieved. This is a significant improvement over the ~ 10 pm sub-image placement accuracy from prior implementations.

[0069] For example, as shown in FIG. 7, the ground truth calculation engine obtains an FOV 802 of a microscope image or sequencing image with fiducials which may include in-tissue fiducials (e.g., metal fiducials) and / or uniquely identifiable fiducials or groups of fiducials. The ground truth calculation engine also obtains a fiducial template 804 indicating the known locations of each of the fiducials on the substrate, the geometries of the fiducials, and the positions of the fiducials with respect to each other.33080 / IP-2902 / PCIP-3041 -PCT

[0070] The ground truth calculation engine may perform a cross-correlation of the fiducials in the FOV 802 with the fiducials in the fiducial template 804 (block 806). For example, the ground truth calculation engine may identify a region of the fiducial template 804 having fiducial geometries, spacing, and / or positions relative to each other which matches the fiducial geometries, spacing, and / or positions relative to each other in the FOV 802. In this manner, the ground truth calculation engine may identify a particular area or cell to which the FOV corresponds.

[0071] At block 808, the ground truth calculation engine may shift the fiducial template 804, such that the identified region of the fiducial template 804 matches the FOV 802. At block 814, the ground truth calculation engine detects the observed locations of the fiducials in the FOV 802, for example as the center locations of the fiducials within the FOV 802. Then, the ground truth calculation engine establishes correspondences between the observed fiducial locations in the FOV 802 and the known locations of the corresponding fiducials in the identified region of the fiducial template 804. Using the correspondences, the ground truth calculation engine estimates an affine transformation by solving for the parameters of the affine transformation based on differences in the observed fiducial locations and the corresponding known fiducial locations, which includes translation, rotation, scaling, and shearing factors (block 810).

[0072] In some implementations, the affine transformation matrix is calculated using unconstrained least squares based on minimizing the sum of the squares of the correspondences between the observed fiducial locations and the known fiducial locations. In other implementations, the affine transformation matrix is calculated using constrained least squares (e.g., if the transformation is within a known range). In still other implementations, the affine transformation matrix is calculated using weighted least squares, where the weights are proportional to confidence levels or correlation peaks reflecting the accuracy of a corresponding fiducial detection.

[0073] Then the ground truth calculation engine applies the affine transformation matrix to the FOV 802 (block 812). In some implementations, the ground truth calculation engine adjusts the position, orientation, and / or scale of the FOV 802 to register the FOV 802 to the ground truth coordinate system.

[0074] At block 816, the ground truth calculation engine detects the locations of the fiducials in the FOV 802 after the transformation, for example as the center locations of the fiducials within the FOV 802. Then the ground truth calculation engine compares the locations of the33080 / IP-2902 / PCIP-3041 -PCT fiducials in the FOV 802 after the transformation to the observed locations of the fiducials in the FOV 802 before the transformation to compute the registration error (block 818).

[0075] While the algorithm 800 in FIG. 7 is shown with respect to FOVs of a microscope image to register the microscope image to a ground truth coordinate system, the algorithm 800 may also be performed with sequencing tiles of a spatial transcript map to register the spatial transcript map to the ground truth coordinate system.

[0076] FIG. 8 depicts another example algorithm 900 for the ground truth calculation engine to register a microscope image or spatial transcript map to a ground truth coordinate system and perform an error assessment. The algorithm 900 is similar to the algorithm 800 shown in FIG. 7.

[0077] As mentioned above, the fiducials described herein may be referred to as “metal- patterned” fiducials, but they may also include nanowell fiducials, or any other suitable fiducials using nanoimprint lithography (NIL). Substrates which include multiple types of fiducials may be referred to as “hybrid substrates."

[0078] An example hybrid substrate is illustrated in FIG. 9. As shown in FIG. 9, a first substrate 1000 includes NIL fiducials placed along a top, middle, and bottom row of the first substrate 1000. The first substrate 1000 also includes metal fiducials placed in the tissue area of the first substrate 1000 with a gap in between regions of metal fiducials so the metal fiducials partially overlap with the NIL fiducials. Two tissues 1002, 1004 are placed on the first substrate 1000, where the metal fiducials are underneath the tissue samples 1002, 1004 and the NIL fiducials are outside of the tissue area for the tissue samples 1002, 1004.

[0079] A second substrate 1050 includes NIL fiducials placed along the left half of the top row of the second substrate 1050, the right half of the bottom row of the second substrate 1050, and rightmost column of the second substrate 1050. The second substrate 1050 also includes metal fiducials placed in the tissue area of the second substrate 1050 without a gap in between regions of metal fiducials so the metal fiducials fully overlap with the NIL fiducials. Two overlapping tissues 1052, 1054 are placed on the second substrate 1050, where the metal fiducials are underneath the tissue samples 1052, 1054 and the NIL fiducials are outside of the tissue area for the tissue samples 1052, 1054.

[0080] FIG. 10 illustrates a detailed view 1100 of the first substrate 1000 shown in FIG. 9. As shown in FIG. 10, the substrate 1000 includes NIL fiducials and metal fiducials. The metal fiducials placed in the tissue area may be referred to as “tissue metal fiducials" or “in-tissue33080 / IP-2902 / PCIP-3041 -PCT fiducials,” and the metal fiducials placed outside of the tissue area, for example along the edges of the substrate 1000, may be referred to as “MarginB metal fiducials” or “MarginB fiducials.”

[0081] Metal fiducials can be detected underneath a tissue sample, whereas other fiducials such as NIL fiducials are more difficult to detect underneath the tissue sample. This allows for the substrate to include additional fiducials along with the NIL fiducials which are placed outside of the tissue area. Accordingly, the ground truth calculation engine may utilize the known locations of in-tissue fiducials as well as the known locations of fiducials placed outside of the tissue area to register the FOVs and sequencing tiles to ground truth. This results in a more accurate registration compared to alternative systems which cannot detect fiducials underneath a tissue sample and solely rely on fiducials placed outside of the tissue area. Such alternative systems may be unable to detect and correct errors in the observed locations within FOVs and sequencing tiles near the middle of a tissue sample, which may not include fiducials placed outside of the tissue area. Moreover, such alternative systems may be more inaccurate at observed locations which are far away from a fiducial placed outside of the tissue area.

[0082] By including in-tissue fiducials which are visible underneath a tissue sample in both FOVs of a microscope image and a sequencing tiles, the fiducial registration system may detect additional fiducials with a higher fiducial detection density, resulting in more accurate transformations and improved registration to ground truth.

[0083] FIG.11 illustrates examples of raw (before correction) and residual (after correction) distortion on two Keyence BZ-X microscopes. Image 1102 shows the raw distortion and image 1 104 the residual distortion on the BZX 800 microscope. Images 1106 shows the raw distortion and image 1 108 the residual distortion on the BZX 710 microscope. The examples demonstrated accuracy measurement on FOV distortion and significant distortion reduction on the same FOV after distortion correction. With the significantly reduced distortion level, image stitching with 10% overlap or even less could potentially be enabled, and the stitched tissue image after FOV distortion correction generated from subsequent full functional assay workflow will have high fidelity within single-cell level precision on histology information.

[0084] FIGs. 12A-12C illustrate other examples of distortion correction responsive to the different patterns of fiducials appearing in captured images of the optical target from various microscopes. FIG. 12A illustrates examples of raw distortion for each of four different microscopes, and FIG. 12B illustrates the corresponding residual distortion for each of the four different microscopes. FIG. 12C illustrates an example of raw and residual distortion on a Nikon Ti2-E microscope. Image 1202 shows the raw distortion, and image 1204 shows the residual33080 / IP-2902 / PCIP-3041 -PCT distortion on the Nikon Ti2-E microscope. In image 1202, the raw distortion on the Nikon Ti2-E microscope is already quite low, and after distortion correction, as shown in image 1204, the distortion is still noticeably reduced. In other words, the method of distortion correction disclosed herein can be used on a microscopes with a wide range of raw distortions and significantly reduces distortion in cases both with a relatively high raw distortion (e.g., the Leica DMi8 microscope illustrated in FIGs. 12A and 12B) and with a relatively low raw distortion (e.g., the Nikon Ti2-E microscope in FIG. 12C).

[0085] The distortion measurement using different DC tile patterned fiducial regions has shown that the optical target establishes a robust and repeatable performance on the same microscopes, and distinguishable on different microscopes, even for microscopes of the same brand with the same model / design specs (distortion profiles for a Keyence microscope shown in FIG.13). FIG.14 shows a plot of registration score versus pixel error size for an example pixel size of 0.380 urn, showing that the present invention can accurately measure pixel size for the FOV being tested in order to support FOVs of various pixel sizes.

[0086] Tighter distribution of the distortion profile from a specific microscope in general results in better image quality and overall healthier microscope lens optical status. This feature enables recognition and continuous tracking of distortion and image quality profile on different microscope lenses. Combined with focus module and shading uniformity, they build a comprehensive microscope imager profile of optical health and image QC profile for any third- party microscope that is being used to generate brightfield stained tissue histology images towards spatial transcriptomics applications, especially for those application that require singlecell resolution.

[0087] In various examples, the evaluation of image quality, including focus (FWHM and Brenner), signal to noise ratio (SNR), contrast, tilt, and depth of field on different microscopes with varied FOV sizes and pixel sizes was achieved by through-focus Z-stack of the DC tile or the neighboring pinhole tile. The shape and sharpness of Z-stack images of the sampled pinhole arrays from either tile can be used to determine the best focus, tilt of the slide (and thus the stage), and usable depth of field for that microscope lens used to generate the images. The shading uniformity for the microscope FOV under specific illumination condition can also be measured by generating the signal heatmap with a rolling window average of 20-50 pixels from a transparent region / blank tile with no metal coating. These design metrics have been demonstrated on other sequencing platforms (e.g. U.S. Patent No. 10,005,083B2), but it is the33080 / IP-2902 / PCIP-3041 -PCT first time they have been established towards external microscope image calibration and QC on a spatial transcriptomics platform.

[0088] FIG.15 illustrates an example distortion estimation workflow 1500 in accordance with an example. At step 1502, an in-focus image is provided to the workflow system. The in-focus image may be obtained using the optical target and techniques illustrated in FIGs 2A-2E.

[0089] At step 1504, the pixel size for the provided image is estimate in pm. If the estimation fails such that the pixel size cannot be determined, the workflow is terminated due to poor image quality at step 1512. If the estimation is successful, the workflow proceeds to step 1506.

[0090] At step 1506, the best ring fiducial size for the provided image is determined and the affine transform is applied. If registration of the provided image to the fiducials fails, the workflow is terminated due to poor image quality at step 1512. If the registration is successful, the workflow proceeds to step 1508.

[0091] At step 1508, image check and pre-processing is performed. If the registered fiducials are too close to the edge of provided image or if the signal-to-background ratio (SBR) is low than a particular threshold, the workflow is terminated due to poor image quality at step 1512. If the registration is successful, the workflow proceeds to step 1510.

[0092] At step 1510, distortion estimation is performed. If the distortion estimation fails, e.g., if the error on the polynomial fit to data is larger than a particular threshold, the workflow is terminated due to poor fit at step 1514. If the distortion estimation is successful, then distortion correction may be performed on the image according to aspects disclosed herein.

[0093] For example, the system obtains the tissue image or sequencing image with the intissue fiducials. In some implementations, the system filters noise from the tissue or sequencing image. The system may also perform a cross-correlation of the fiducials with a fiducial template. The system may then align the tissue or sequencing image based on the cross-correlation by shifting the coordinates of the tissue or sequencing image in accordance with the comparison to the fiducial template.

[0094] In any event, the system compares the locations of the in-tissue fiducials in the tissue or sequencing image to their known physical locations to identify differences between the locations of the in-tissue fiducials in the tissue or sequencing image and their known physical locations. Then the system computes a transform based on these differences and generate an affine transformation matrix to translate, rotate, or scale the tissue or sequencing image. The system then applies the affine transformation matrix to the tissue or sequencing image to33080 / IP-2902 / PCIP-3041 -PCT generate a coordinate system for the tissue or sequencing image indicating the physical locations within the tissue or sequencing image.

[0095] In some implementations, the system detects the locations of the fiducial centers after performing a cross-correlation based alignment and detects the locations of the fiducial centers after applying the affine transformation matrix and compares these locations. The system may then compute the registration error of the imaging software based on this analysis.

[0096] In some implementations, one repeating unit of the substrate matches the size of one FOV of the microscope. Each FOV may include 6x4 patterned fiducials which provides high density of fiducials to accurately measure the distortion error. The fiducials may be rings and may mimic conventional “bullseye” fiducials that can be used in the sequencing tile registration.

[0097] The high contrast of a metal fiducial against tissue provides good visibility and precise detection. The metal fiducial template also allows the system to determine the effect of using edge fiducials (vs all fiducials) on the registration error.

[0098] In some scenarios, the error may be generated from stitching FOVs together to generate the complete view of the tissue image. To account for this, the system can evaluate the registration error with FOV distortion correction followed by re-stitching the FOV images.

[0099] FIG. 16 depicts a network diagram of an exemplary network environment 1600 which may be used to perform the methods described herein. The network environment includes an analysis system 1610 (e.g., a transcriptomics analyzer), a microscope-based sequencing system 1620, and an image database 1630 all communicatively connected by communication network 1640. Communication network 1640 may be any suitable network type, such as a local area network (“LAN”) or a wide area network (“WAN”), e.g., the Internet.

[0100] In some aspects, the analysis system 1610 may receive a microscope image 1642 of a tissue sample captured by a microscope. The microscope image 1642 may be microscope image 1622 provided by the microscope-based sequencing system 1620, or the microscope image 1642 may be microscope image 1632 stored in the image database 1630.

[0101] In some aspects, the analysis system 1610 may determine profile data 1612 for the microscope which was used to capture the received microscope image 1642. Microscope profile data 1612 accessible by the analysis system 1610 may include data for many different microscopes, including the microscope of the microscope-based sequencing system 1620 and any microscopes used to collect microscope images stored in the image database 1630. The profile data 1612 may comprise distortion correction data, illumination intensity correction data,33080 / IP-2902 / PCIP-3041 -PCT and focusing correction data generated from analyzing a plurality of test images captured by the microscope of an optical target comprising a blank region for determining the illumination intensity correction data, a focusing zone region for determining the focusing correction data, and a distortion control region for determining the distortion correction data. The optical target may be similar to optical target slide 200 shown in FIG. 2A.

[0102] In some aspects, the analysis system 1610 may provide the microscope image 1642 to an image correction process 1614 to analyze the microscope image 1642 based on the profile data 1612 and to generate a corrected microscope image 1616. The image correction process 1614 may include distortion correction, focus / image quality evaluation, and / or shading uniformity correction.

[0103] In some aspects, the analysis system 1610 may receive sequencing images 1644 of the tissue sample and provide the sequencing images to a spatial transcriptomics process 1618. The sequencing images 1644 may be sequencing images 1624 provided by the microscopebased sequencing system 1620, or the sequencing images 1644 may be sequencing images 1634 stored in the image database 1630.

[0104] In some aspects, the analysis system 1610 may generate, via the spatial transcriptomics process, a spatial transcriptomic heatmap image 1646 and a spatial transcriptomics image report 1648 comprising an overlay of the spatial transcriptomic heatmap image 1646 and the corrected microscope image 1616.

[0105] Thus, as shown, the present techniques provide on-market spatial transcriptomics platforms that address distortion and effectively correct distortion for microscope histology images supporting microscope QC for different vendors with different models and specs.

[0106] The optical targets and techniques herein may be integrated into various spatial transcriptomic workflows, as well as other medical imaging workflows where registration between images for tissue-level, cellular-level, and sub-cellular level registration accuracy is needed.

[0107] The optical targets and techniques herein may be integrated into various model systems that integrate metal -patterned fiducials into an image alignment workflow, for example to better characterize the distortion error and precisely register and stitch the images to the reference coordinates, which that contains the theoretical XY location for each sub-image.

[0108] Aspects of the techniques described in the present disclosure may include any of the following aspects, either alone or in combination:33080 / IP-2902 / PCIP-3041 -PCT

[0109] 1 . A method for compensating for quality variation in an imager capturing microscope images or sequencing images associated with a tissue, the method comprising: receiving a plurality of images of an optical target, captured by an imager, the optical target comprising a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region, wherein the plurality of images comprises an image of each different area of the distortion control region; providing one or more of the images of the different areas of the distortion control region to a distortion correction process configured to perform a distortion correction based on locations of one or more of the patterned fiducial regions; generating profile data for the imager, the profile data comprising the distortion correction data from the distortion correction process; and storing the quality variation profile data for use in correcting subsequent microscope images or subsequent sequencing images received from the imager.

[0110] 2. The method of aspect 1 , the optical target comprises a blank region and a focusing zone region, and wherein the plurality images of the optical target comprise at an image of the blank region and at least one image of the focusing zone region.

[0111] 3. The method of aspect 2, further comprising: providing the image of the blank region to an illumination intensity correction process, wherein generating the profile data for the imager further comprises determining, by the illumination intensity correction process, illumination intensity correction data and storing the illumination intensity correction data in the profile data.

[0112] 4. The method of aspect 2 or aspect 3, further comprising: providing the at least one image of the focusing zone region to a focusing process, wherein generating the profile data for the imager further comprises determining, by the focusing process, focusing correction data and storing the focusing correction data in the profile data.

[0113] 5. The method of aspect 4, wherein the at least one image of the focusing zone comprises a Z-stack of images of the focusing zone.

[0114] 6. The method of any of aspects 1 -5, wherein the distortion correction process comprises one or more of a radial (or other 2-dimensional) distortion correction, a scale correction, a rotation correction, a translation correction, a skew correction, a projective correction, and an affine correction.

[0115] 7. The method of any of aspects 1 -6, wherein the plurality of patterned fiducial regions are concentric and each patterned fiducial region spans a different area size.33080 / IP-2902 / PCIP-3041 -PCT

[0116] 8. The method of any of aspects 1 -7, wherein the plurality of patterned fiducial regions each have the same geometric shape.

[0117] 9. The method of any of aspects 1 -8, wherein each patterned fiducial region comprises a plurality of fiducials each formed as a pinhole array.

[0118] 10. The method of aspect 9, wherein each of the fiducials for each patterned fiducial region differ in size from the fiducials in each other patterned fiducial region.

[0119] 11 . The method of aspect 9 or aspect 10, wherein each of the plurality of patterned fiducial regions have the fiducials that differ in size from each other of the plurality of patterned fiducial regions.

[0120] 12. An optical target comprising: a metal patterned substrate having a blank region, a focusing zone region comprising a plurality of focusing features, and a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region.

[0121] 13. The optical target of aspect 12, wherein the blank region, the focusing zone region, and the distortion control region are confined to an imaging space of between 450 pm x 450 pm and 2200 pm x 2200 pm or a rectangular space of 450pm x 2200pm of smaller .

[0122] 14. The optical target of aspect 12 or aspect 13, wherein the plurality of patterned fiducial regions are concentric.

[0123] 15. The optical target of any of aspects 12-14, wherein the plurality of patterned fiducial regions differ in area size.

[0124] 16. The optical target of any of aspects 12-14, wherein the plurality of patterned fiducial regions have the same geometric shape.

[0125] 17. The optical target of any of aspects 12-16, wherein each patterned fiducial region comprises a plurality of fiducials each formed as a pinhole array.

[0126] 18. The optical target of aspect 17, wherein each of the fiducials for each patterned fiducial region differ in size from the fiducials in each other patterned fiducial region.

[0127] 19. The optical target of aspect 17, wherein the pinhole array has a diameter of 1 pm or less.33080 / IP-2902 / PCIP-3041 -PCT

[0128] 20. The optical target of any of aspects 17-19, wherein the distortion control region is configured to allow for correcting microscope image resolutions from 0.15 pm / pixel to 0.5 pm / pixel.

[0129] 21 . The optical target of any of aspects 12-20, where the focusing zone region comprises a pinhole as the plurality of focusing features.

[0130] 22. A method for performing spatial transcriptomics, the method comprising: receiving a microscope image of a tissue sample captured by a microscope; determining profile data for the microscope, the profile data comprising distortion correction data, illumination intensity correction data, and focusing correction data generated from analyzing a plurality of test images captured by the microscope of an optical target comprising a blank region for determining the illumination intensity correction data, a focusing zone region for determining the focusing correction data, and a distortion control region for determining the distortion correction data; providing the microscope image to an image correction process to analyze the microscope image based on the profile data and to generate a corrected microscope image; receiving sequencing images of the tissue sample and providing the sequencing images to a spatial transcriptomics process; generating, via the spatial transcriptomics process, a spatial transcriptomic heatmap image; and generating a spatial transcriptomics image report comprising an overlay of the spatial transcriptomic heatmap image and the corrected microscope image.

[0131] 23. The method of aspect 22, wherein the corrected microscope image has a raw distortion of between and including 6pm - 20pm across the corrected microscope image.

[0132] 24. The method of aspect 22 or aspect 23, wherein the corrected microscope image has a distortion correction of less than and including 0.5 pm across the corrected microscope image.

[0133] 25. The method of any of aspects 22-24, wherein the blank region, the focusing zone region, and the distortion control region are confined to an imaging space of between 450pm x 450pm and 2200 pm x 2200 pm or a rectangular space of 450 pm x 2200 pm or smaller.ADDITIONAL CONSIDERATIONS

[0134] Although the disclosure herein sets forth a detailed description of numerous different implementations, it should be understood that the legal scope of the description is defined by the words of the claims set forth at the end of this patent and equivalents. The detailed description is to be construed as exemplary only and does not describe every possible33080 / IP-2902 / PCIP-3041 -PCT implementation since describing every possible implementation would be impractical.Numerous alternative implementations may be implemented, using either current technology or technology developed after the filing date of this patent, which would still fall within the scope of the claims.

[0135] The following additional considerations apply to the foregoing discussion. Throughout this specification, plural instances may implement components, operations, or structures described as a single instance. Although individual operations of one or more methods are illustrated and described as separate operations, one or more of the individual operations may be performed concurrently, and nothing requires that the operations be performed in the order illustrated. Structures and functionality presented as separate components in example configurations may be implemented as a combined structure or component. Similarly, structures and functionality presented as a single component may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of the subject matter herein.

[0136] Additionally, certain implementations are described herein as including logic or a number of routines, subroutines, applications, or instructions. These may constitute either software (e.g., code embodied on a machine-readable medium or in a transmission signal) or hardware. In hardware, the routines, etc., are tangible units capable of performing certain operations and may be configured or arranged in a certain manner. In example implementations, one or more computer systems (e.g., a standalone, client or server computer system) or one or more hardware modules of a computer system (e.g., a processor or a group of processors) may be configured by software (e.g., an application or application portion) as a hardware module that operates to perform certain operations as described herein.

[0137] The various operations of example methods described herein may be performed, at least partially, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors may constitute processor-implemented modules that operate to perform one or more operations or functions. The modules referred to herein may, in some example implementations, comprise processor-implemented modules.

[0138] Similarly, the methods or routines described herein may be at least partially processor implemented. For example, at least some of the operations of a method may be performed by one or more processors or processor-implemented hardware modules. The performance of certain of the operations may be distributed among the one or more processors,33080 / IP-2902 / PCIP-3041 -PCT not only residing within a single machine, but deployed across a number of machines. In some example implementations, the processor or processors may be located in a single location, while in other implementations the processors may be distributed across a number of locations.

[0139] The performance of certain of the operations may be distributed among the one or more processors, not only residing within a single machine, but deployed across a number of machines. In some example implementations, the one or more processors or processor- implemented modules may be located in a single geographic location (e.g., within a home environment, an office environment, or a server farm). In other implementations, the one or more processors or processor-implemented modules may be distributed across a number of geographic locations.

[0140] This detailed description is to be construed as exemplary only and does not describe every possible implementation, as describing every possible implementation would be impractical, if not impossible. A person of ordinary skill in the art may implement numerous alternate implementations, using either current technology or technology developed after the filing date of this application.

[0141] Those of ordinary skill in the art will recognize that a wide variety of modifications, alterations, and combinations may be made with respect to the above described implementations without departing from the scope of the invention, and that such modifications, alterations, and combinations are to be viewed as being within the ambit of the inventive concept.

[0142] The patent claims at the end of this patent application are not intended to be construed under 35 U.S.C. § 1 12(f) unless traditional means-plus-function language is expressly recited, such as “means for” or “step for” language being explicitly recited in the claim(s). The systems and methods described herein are directed to an improvement to computer functionality and improve the functioning of conventional computers.

Claims

33080 / IP-2902 / PCIP-3041 -PCTWhat is Claimed is:1 . A method for compensating for quality variation in an imager capturing microscope images or sequencing images associated with a tissue, the method comprising: receiving a plurality of images of an optical target, captured by an imager, the optical target comprising a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region, wherein the plurality of images comprises an image of each different area of the distortion control region; providing one or more of the images of the different areas of the distortion control region to a distortion correction process configured to perform a distortion correction based on locations of one or more of the patterned fiducial regions; generating profile data for the imager, the profile data comprising the distortion correction data from the distortion correction process; and storing the quality variation profile data for use in correcting subsequent microscope images or subsequent sequencing images received from the imager.

2. The method of claim 1 , the optical target comprises a blank region and a focusing zone region, and wherein the plurality images of the optical target comprise at an image of the blank region and at least one image of the focusing zone region.

3. The method of claim 2, further comprising: providing the image of the blank region to an illumination intensity correction process, wherein generating the profile data for the imager further comprises determining, by the illumination intensity correction process, illumination intensity correction data and storing the illumination intensity correction data in the profile data.

4. The method of claim 2, further comprising: providing the at least one image of the focusing zone region to a focusing process, wherein generating the profile data for the imager further comprises determining, by the focusing process, focusing correction data and storing the focusing correction data in the profile data.

5. The method of claim 4, wherein the at least one image of the focusing zone comprises a Z-stack of images of the focusing zone.33080 / IP-2902 / PCIP-3041 -PCT6. The method of claim 1 , wherein the distortion correction process comprises one or more of a radial (or other 2-dimensional) distortion correction, a scale correction, a rotation correction, a translation correction, a skew correction, a projective correction, and an affine correction.

7. The method of claim 1 , wherein the plurality of patterned fiducial regions are concentric and each patterned fiducial region spans a different area size.

8. The method of claim 1 , wherein the plurality of patterned fiducial regions each have the same geometric shape.

9. The method of claim 1 , wherein each patterned fiducial region comprises a plurality of fiducials each formed as a pinhole array.

10. The method of claim 9, wherein each of the fiducials for each patterned fiducial region differ in size from the fiducials in each other patterned fiducial region.1 1 . The method of claim 9, wherein each of the plurality of patterned fiducial regions have the fiducials that differ in size from each other of the plurality of patterned fiducial regions.

12. An optical target comprising: a metal patterned substrate having a blank region, a focusing zone region comprising a plurality of focusing features, and a distortion control region formed of a plurality of patterned fiducial regions each defining a different area of the distortion control region.

13. The optical target of claim 12, wherein the blank region, the focusing zone region, and the distortion control region are confined to an imaging space of between 450 pm x 450 pm and 2200 pm x 2200 pm or a rectangular space of 450 pm x 2200 pm of smaller .

14. The optical target of claim 12, wherein the plurality of patterned fiducial regions are concentric.

15. The optical target of claim 12, wherein the plurality of patterned fiducial regions differ in area size.33080 / IP-2902 / PCIP-3041 -PCT16. The optical target of claim 12, wherein the plurality of patterned fiducial regions have the same geometric shape.

17. The optical target of claim 12, wherein each patterned fiducial region comprises a plurality of fiducials each formed as a pinhole array.

18. The optical target of claim 17, wherein each of the fiducials for each patterned fiducial region differ in size from the fiducials in each other patterned fiducial region.

19. The optical target of claim 17, wherein the pinhole array has a diameter of 1 pm or less.

20. The optical target of claim 17, wherein the distortion control region is configured to allow for correcting microscope image resolutions from 0.15 pm / pixel to 0.5 pm / pixel.21 . The optical target of claim 12, where the focusing zone region comprises a pinhole as the plurality of focusing features.

22. A method for performing spatial transcriptomics, the method comprising: receiving a microscope image of a tissue sample captured by a microscope; determining profile data for the microscope, the profile data comprising distortion correction data, illumination intensity correction data, and focusing correction data generated from analyzing a plurality of test images captured by the microscope of an optical target comprising a blank region for determining the illumination intensity correction data, a focusing zone region for determining the focusing correction data, and a distortion control region for determining the distortion correction data; providing the microscope image to an image correction process to analyze the microscope image based on the profile data and to generate a corrected microscope image; receiving sequencing images of the tissue sample and providing the sequencing images to a spatial transcriptomics process; generating, via the spatial transcriptomics process, a spatial transcriptomic heatmap image; and generating a spatial transcriptomics image report comprising an overlay of the spatial transcriptomic heatmap image and the corrected microscope image.33080 / IP-2902 / PC IP-3041 -PCT23. The method of claim 22, wherein the corrected microscope image has a raw distortion of between and including 6 pm - 20 pm across the corrected microscope image.

24. The method of claim 22, wherein the corrected microscope image has a distortion correction of less than and including 0.5 pm across the corrected microscope image.

25. The method of claim 22, wherein the blank region, the focusing zone region, and the distortion control region are confined to an imaging space of between 450 pm x 450 pm and 2200 pm x 2200 pm or a rectangular space of 450 pm x 2200 pm or smaller.

Images