Methods and systems for histological characterization of the liver

Abstract

Methods and systems for histological characterization of the liver are disclosed. The methods are used for detection and characterization of the portal tract. In particular, the invention relates to methods and systems for determining indices of the portal tract, anatomic microcomponents, and related areas in liver histological sections stained with a plurality of different stains.

Description

Technical Field

[0001] This invention relates to methods for the detection and characterization of the portal region. Specifically, this invention relates to methods and systems for determining indicators of the portal region, anatomical micro-components, and related regions in liver histological sections with various staining methods. Background Technology

[0002] The portal tract (also known as the portal triad) is an aggregation of anatomical micro-components present in the liver. These micro-components include the hepatic artery, portal vein, and bile ducts, all surrounded by connective tissue. These micro-components facilitate the flow of nutrient-rich blood (portal vein), oxygenated blood (hepatic artery), and bile (bile ducts) within the liver.

[0003] Traditionally, based on blood supply and the distance between the portal area and its nearest central vein, the liver microanatomy is divided into three acinar zones with distinct metabolic biological characteristics. These acinar zones are located within acini, which are rhomboid regions within each liver lobule. The portal area is located at the corner of each acinus. The zone furthest from the central vein is called zone one and receives the most oxygenated blood and nutrients. The zone closest to the central vein is called zone three and receives the least oxygenated blood and nutrients.

[0004] The location of certain disease features within each acinar region can indicate the severity of the disease. This is accompanied by the presence of hepatocellular damage, inflammation, fibrosis, and fatty deposits found in the three acinar regions [S. Hübscher, Histological assessment of non-alcoholic fatty liver disease, Histopathology, 49 (2006) 450-465].

[0005] The portal region is an important aggregate that needs to be identified in liver histological sections because its structure and location inform the pathologist's assessment of the patient's health. For example, in metabolic-associated steatohepatitis (MASH), portal inflammation is associated with poorer clinical outcomes. Portal inflammation is defined as inflammation confined to the portal region of the liver tissue.

[0006] Whether in clinical trials or routine diagnostic settings, gate canal detection is a crucial part of biopsy sample quality control in liver pathology. Among other criteria (such as minimum biopsy core length and lack of folds), a biopsy slide must possess a sufficient number of gate canals to be considered suitable for evaluation. The required number varies depending on the environment, but guidelines issued by the British Society of Gastroenterology, the Royal College of Pathologists, and the American Association for the Study of Liver Diseases stipulate that 6-11 portal areas per sample are sufficient, while some studies suggest that the presence of more than 11 complete portal areas constitutes a qualified sample [A. Boyd, O. Cain, A. Chauhan, GJ Webb, Medical liver biopsy:background, indications, procedure and histopathology, Frontlinegastroenterology, 11 (2020) 40-47; E. Cholongitas, M. Senzolo, R. Standish, L. Marelli, A. Quaglia, D. Patch, AP Dhillon, AK Burroughs, A systematic review of the quality of liver biopsy specimens, American journal of clinicalpathology, 125 (2006) 710-721; DC Rockey, SH Caldwell, ZD Goodman, RC Nelson, AD Smith, Liver biopsy, Hepatology, 49 (2009) 1017-1044; ARCrawford, X.-Z. Lin, JM Crawford, The normal adult human liver biopsy: aquantitative reference standard, Hepatology, 28 (1998) 323-331].

[0007] Six complete portal zones can be found per linear centimeter of liver tissue [Crawford AR, Lin XZ, Crawford JM. The normal adult human liver biopsy: a quantitative reference standard. Hepatology. 1998;28(2):323-331]. The longer the liver biopsy tissue, the more portal zones are spanned throughout the specimen, and therefore the more reliable the grading and staging of liver disease progression. Many trials mandate the rejection of biopsy samples that do not meet these criteria.

[0008] Detection and identification of gate regions are challenging because they vary greatly in size, structure, shape, and overall appearance. The location and angle of biopsy needle insertion also affect the appearance of gate regions on the slide.

[0009] Most liver diseases affect the distribution and appearance of portal areas, and this effect increases with disease severity. This is evident in the development of fibrotic bridging that gradually grows from one portal area to another.

[0010] Fibrosis scores are influenced by the quantity, structure, and location of collagen deposits in liver histological sections stained with connective tissue. Sections with bridging are given higher scores. Bridging is defined as the connection between two or more portal areas via collagen bridging.

[0011] Fibrosis regression is classified using a scale (F0–F4), where F0 represents no fibrosis, F1 represents portal fibrosis without septa (strips of scar tissue), F2 represents portal fibrosis with a few septa, F3 represents many septa but no cirrhosis, and F4 represents cirrhosis.

[0012] Later stages (F3 and F4) of fibrosis regression exhibit rounded and / or interrupted bridging, with or without aligned septal ends, indicating the location of previously visible intact bridging and the re-establishment of hepatocytes in the scarred area. Most fibrous septal matrix shows signs of the liver repair complex, appearing as a dense, compressed, heavily stained, and predominantly acellular matrix [S. Khan, R. Saxena, Regression of hepatic fibrosis and evolution of cirrhosis: a concise review, Advances in anatomic pathology, 28(2021) 408-414].

[0013] Early-stage (F1 and F2) fibrosis regression is characterized by the absence of perisinusoidal / pericellular fibrosis.

[0014] Fibrosis regression is often a biomarker for the efficacy of medications in liver diseases [AJ Sanyal, QM Anstee, M. Trauner, EJ Lawitz, MF Abdelmalek, D. Ding, L. Han, C. Jia, RS Huss, C. Chung, Cirrhosis regression is associated with improved clinical outcomes in patients with nonalcoholic steatohepatitis, Hepatology, 75 (2022) 1235-1246], even when AI-based endpoints are used in clinical trials [R. Loomba, P. Bedossa, K. Grimmer, G. Kemble, EB Martins, W. McCulloch, M. O'Farrell, W.-W. Tsai, J. Cobiella, E. Lawitz, Denifanstat for the treatment of metabolic dysfunction-associated steatohepatitis: a multicentre, double-blind, randomised, placebo-controlled, phase 2b trial, The Lancet] Gastroenterology & Hepatology, (2024)].

[0015] Fibrosis quality is also a potential surrogate endpoint, which has been shown to be an accurate indicator in the treatment of chronic hepatitis B [Y. Sun, W. Chen, S. Chen, X. Wu, X. Zhang, L. Zhang, H. Zhao, M. Xu, Y. Chen, H. Piao, Regression of Liver Fibrosis in Patients on Hepatitis B Therapy Is Associated With Decreased Liver-Related Events, Clinical Gastroenterology and Hepatology, 22 (2024) 591-601. e593].

[0016] The process of defining the boundaries of anatomical micro-fractions present in the liver, or the boundaries of aggregates within micro-fractions (such as portal regions and fibrotic areas), is called segmentation or delineation. This boundary can be represented in various ways, such as as a set of coordinates, with each instance of a micro-fraction or aggregate represented as a polygon. The boundary can also be represented as a mask image, where the pixel dimensions match the pixel dimensions of the original image, and the value of each pixel in the mask determines whether it is inside (foreground) or outside (background) an instance of the structure of interest. Whether in polygon or mask form, delineation can be created using a manual process (e.g., using software with a set of drawing tools) or an automated model (e.g., using a convolutional neural network).

[0017] The segmentation of the portal canal allows for the study of other indicators of disease progression. These include delineating the relationship between lobular and portal inflammation, fat deposition, and the location and extent of fibrosis.

[0018] Areas of elevated leukocyte density are associated with inflammation. A single, continuous tissue region with a density consistently above a certain level is described as an inflammatory focus. Very small structures, such as the nuclei of leukocytes, can be represented either by a mask or by a set of coordinates representing the centers of the individual nuclei. Socha et al. provide an example of how inflammatory focuses can be measured and quantified [P. Socha, E. Shumbayawonda, A. Roy, C. Langford, P. Aljabar, M. Wozniak, S. Chelstowska, E. Jurkiewicz, R. Banerjee, K. Fleming, Quantitative digitalpathology enables automated and quantitative assessment of inflammatory activity in patients with autoimmune hepatitis, Journal of PathologyInformatics, 15 (2024) 100372].

[0019] Measurements of steatosis, inflammation, and fibrosis are typically reported as averages across the entire slice. This excludes measurements of the heterogeneity of the presence of features across the entire slice. However, heterogeneity is a clinically important consideration because different liver diseases exhibit different patterns of portal inflammation, lobular inflammation, and interface hepatitis.

[0020] Several commercially available automated segmentation methods have been applied to liver histopathology with varying performance. Some have used second harmonic generation (SHG) and two-photon excitation fluorescence (TPFE) microscopy to distinguish between portal and lobular regions, and to identify portal and central veins to detect inflammation and quantify fibrosis (Y. Wei, M. Zhang, J. Zhang, X. Teng, Q. Yang, A. Leong, G. Ho, H. Lu, D. Xu, Clinical relevance of ananimal model of non-alcoholic steatohepatitis (NASH) and digital pathology with artificial intelligence (DP-AI) analyses of hepatic fibrosis, Journal of Hepatology, 77 (2022) S679-S680). This method has been applied to assess fibrosis dynamics in MASH patients with bridging fibrosis and fibrosis regression, using the qFibrosis index to evaluate the efficacy of drug treatment or lifestyle intervention (NV Naoumov, DE Kleiner, E. Chng, D. Brees, C. Saravanan, Y. Ren, D. Tai, AJ Sanyal, Digital quantitation of bridging fibrosis and septa reveals changes in natural history and treatment not seen with conventional histology, Liver International, (2024); HY Yuan, XF Tong, YY Ren, YYLi, XL Wang, LL Chen, SD Chen, XZ Jin, XD Wang, G. Targher, Al-based digital pathology provides newer insights into lifestyle intervention-induced fibrosis regression in MASLD: An exploratory study, Liver International, (2024)).

[0021] Another commercially available automated segmentation method assesses MASH / MAFLD in preclinical rodent models of NASH. This automated method relies on morphometric identification and requires Sirius red (PSR) staining to quantify fibrosis, although it is unclear whether the portal area is detected to analyze fibrosis. Hematoxylin-eosin (H&E) sections and F4 / 80 (an antibody marker for macrophages widely used in murine models) are both used to quantify inflammation. However, a 2020 article showed that these methods do not differentiate between portal and lobular inflammation, and immunohistochemical (IHC) staining has since been used significantly more extensively (M. De Rudder, C. Bouzin, M. Nachit, H. Louvegny, G. Vande Velde, Y. Jule, IA Leclercq, Automated computerized image analysis for the user-independent evaluation of disease severity in preclinical models of NAFLD / NASH, Laboratory investigation, 100 (2020) 147-160).

[0022] Vanderbeck et al. from Wisconsin used an independent AI model of portal and lobule to detect lobular inflammation (S. Vanderbeck, J. Bockhorst, D. Kleiner, R. Komorowski, N. Chalasani, S. Gawrieh, Automatic quantification of lobular inflammation and hepatocyte ballooning in nonalcoholic fatty liver disease liver biopsies, Humanpathology, 46 (2015) 767-775). Yu et al. used a similar model to identify and quantify the portal region (H. Yu, N. Sharifai, K. Jiang, F. Wang, G. Teodoro, AB Farris, J. Kong, Artificialintelligence based liver portal tract region identification and quantification with transplant biopsy whole-slide images, Computers in Biology and Medicine, 150 (2022) 106089).

[0023] Another noteworthy model is based on the quantification of fibrosis in the portal region, which is based on the intersection with the septum and the assessment of the fibrotic regions of the partition (M. Noureddin, Z. Goodman, D. Tai, EL Chng, Y. Ren, P. Boudes, H. Shlevin, G. Garcia-Tsao, SA Harrison, NP Chalasani, Machine learning liver histology scores correlate with portal hypertension assessments in nonalcoholic steatohepatitis cirrhosis, Alimentary Pharmacology & Therapeutics, 57 (2023) 409-417).

[0024] However, there is currently no commercial or other solution that can provide an interpretive framework for the allocation of gate region segmentation within samples by various systems. Summary of the Invention

[0025] According to a first aspect of the invention, a method and system for histological characterization of the liver, comprising all the features of claim 1, are provided. Further features of the invention are disclosed in the dependent claims.

[0026] According to one aspect of the present invention, a method for determining liver histological parameters is provided, comprising:

[0027] Segmentation of histological micro-components from biological samples using liver histological images;

[0028] Spatially correlate the micro-components within a preselected proximity to each other into micro-component aggregates, which define histological components or indicators of disease severity; and

[0029] Each of the micro-components within a preselected proximity to each other is associated with the bile duct, portal vein, and hepatic artery in the portal region, such that the confidence level of the aggregates of the micro-components exceeds a threshold.

[0030] The method provides an interpretative framework for the allocation of aggregates containing segmentation of gate regions within a sample. It includes means of correlating spatial analysis with the segmentation process to interpret which known gate region biological elements are present in the resulting aggregate segmentation. Spatial analysis for spatially correlating the micro-components may include the analysis of patterns of micro-components, patterns of aggregates, or both.

[0031] One possible pattern is a hexagonal arrangement of liver lobules. In this pattern, each of the six outer corners of each roughly hexagonal liver unit roughly contains a portal canal. The hexagonal arrangement of liver lobules presents as a hexagonal honeycomb pattern. The center of each hexagon roughly corresponds to the central vein.

[0032] Another pattern may include nested rhombuses representing acinar regions. Smaller rhombuses may be nested within larger rhombuses. All rhombuses may have a first pair of diagonals and a second pair of diagonals. All rhombuses may share the first pair of diagonals. A portal canal may be located at each of the two diagonals of the first pair. A central vein may be located at each of the two diagonals of the second pair of the largest rhombus. However, smaller rhombuses may not have diagonals sharing a central vein, as they are nested within the largest rhombus. The nested rhombuses may correspond to acinar regions.

[0033] Another possible pattern is an aggregation of at least one portal vein, hepatic artery, and bile duct within a certain distance from each other in connective tissue.

[0034] Detected and identified micro-components or aggregates, or both, can be spatially correlated to determine whether they fit one or more patterns. This fit can be determined by the relative orientation and distance to previously detected and identified micro-components and aggregates. These patterns can be known to place specific micro-components and aggregates at corresponding locations within a range of distances and angles relative to one or more micro-components in that pattern. Histological characterization methods can be improved as the analysis of histological sections progresses, becoming faster and more accurate in detecting each subsequent micro-component or aggregate based on the patterns filled by the detected micro-components or aggregates.

[0035] Micro-components may include: tissue foreground, lobules, portal area, connective tissue, bile ducts, arteries, veins, fat globules, periportal space, sinusoidal space, hepatic border plate, hepatocyte type or combinations thereof.

[0036] The confidence level of an aggregate can be determined based on the presence of colocalized elements, such as colocalized micro-components or micro-components within a preselected proximity of the aggregate or aggregate segmentation. Proximity can be determined by the shortest distance from the segmentation edge to the segmentation edge of the micro-component.

[0037] Spatial domains or patterns can be derived from portal and component features using geometric, cartographic, or morphological operations, wherein the spatial domains or patterns include at least one of the lobular regions, hepatic partitions, or interface regions of the liver.

[0038] The detection and identification of each micro-component or its aggregates, patterns, or spatial domains can be used to derive liver disease characteristics present within liver lobules and adjacent elements.

[0039] The confident detection and identification of micro-components and their aggregates in biological samples as gate regions, or the detection and identification of their spatial domains or patterns, can be used to provide measurement information or indicators.

[0040] Micro-components can be spatially correlated relative to each other or relative to aggregates or both using cartographic analysis, morphological analysis, or both.

[0041] The presence or absence of spatially associated micro-components, their aggregates, or patterns can influence an indicator or confidence score used to determine the likelihood that a gate region is "true." In this way, candidate gate regions can be identified as true gate regions. The use of micro-components, aggregates of micro-components (such as gate regions), or the use of spatial domains can isolate them from other spatial features of the biological sample on a histological section. Gate regions can be quantified to assess sample adequacy.

[0042] Preferably, the location of inflammatory cells, balloon cells, fibrosis, fat globule features, or connective tissue is integrated with the lobular region and portal canal to provide an indicator of the spatial differentiation of the number and density of inflammatory cells within and outside the portal canal, or to identify bridging fibrosis between portal canals.

[0043] Segmentation models can be used to identify the component features of gate regions. Micro-components, aggregates thereof, or parts thereof can be segmented using neural networks, convolutional neural networks, Transformer models, including visual Transformers, or combinations of multiple such networks trained for this purpose. They can be segmented using neural networks containing a base segmentation model, or autoencoders or Transformers capable of learning and generating embedding vectors of arbitrary image features. Micro-components, aggregates thereof, or parts thereof can also be segmented using non-neural network-based image processing techniques such as thresholding, pixel classification, filtering, or feature detection.

[0044] Micro-components, aggregates thereof, or portions thereof can be roughly segmented using patch-based classification, where smaller patches are extracted from a larger image and classified based on whether the patch contains the presence of each component's features. Micro-components, aggregates thereof, or portions thereof (such as the cells they contain) can be identified as aggregates of bounding boxes, centroids, or other coordinates, and these aggregate descriptors can be spatially correlated.

[0045] Micro-components, aggregates thereof, or portions of micro-components (such as cells) can be represented as location-based probabilities of the presence of that micro-component, aggregate, or portion at a corresponding location on a probability map. Spatial association of component features is performed by the geometry derived from thresholding the probability map at a certain value. Micro-components or aggregates thereof can be identified by segmenting the cell types they comprise using a cell segmentation or detection model trained to predict cell types. Micro-components or aggregates thereof that are connective tissue or within connective tissue can be identified as connective tissue cells or micro-components or aggregates within each other's proximity. Micro-components or portions thereof can be detected or identified by the cell types they contain. For example, bile ducts can be identified by bile duct epithelial cells within each other's proximity in micro-components.

[0046] The portal regions and hepatic lobular regions can be delineated using logical predicates, including union, intersection, or difference, by combining the component features of the portal regions with the foreground features. Interface regions can be delineated between the portal regions and hepatic lobular regions through constructive and destructive geometric operations, including features of buffering, dilation, or erosion. For example, a midline can be found between two portal regions and the lobular region closest to one of the two portal regions, and further subdivided by the radial distance from one boundary of a portal region to the other. For the partitions to be analyzed in liver histological images, delineation can be achieved by subdividing the portal regions and hepatic lobular regions.

[0047] The gate region model can be used to distinguish between the gate region, leaflet region, and interface region by generating descriptive statistics.

[0048] The distribution of inflammation in the portal canal, lobular region, or interface can be expressed by at least one of the following: the number of inflammatory foci, the average area of ​​the foci, the density of inflammatory cells in the portal canal, the density of inflammatory cells in the interface, or the density of inflammatory cells in the lobular region.

[0049] Preferred spatial locations of neutrophils, eosinophils, lymphocytes (including T cells and B cells), plasma cells, macrophages (including Kupffer cells), or other inflammatory cell types relative to the portal region, lobular region, and other liver regions can be identified. Clusters of cooperating cells defined by statistical frequency, nearest neighbor, "bag of words," or other techniques can be used to determine their location or abundance relative to the portal region, lobular region, or interface region.

[0050] The portal region can be evaluated, or the portal region can be combined with micro-components in connective tissue, or connective tissue, or both, to identify bridging fibrosis between portal regions. The shape, length, and other morphological characteristics of connective tissue or micro-components in connective tissue, or both, relative to the portal region can be assessed to evaluate the formation of collagen bridging.

[0051] The biological samples can be used to assess changes in portal areas, lobular regions, interfaces, or connective tissue over time, including assessing the reduction or increase of inflammation, fibrosis, and scarring. Biological samples in liver histological images can be longitudinal samples. Over time, scarring or fibrosis can occur, creating bridging between portal areas. The length or width, or both, of the bridging fibrosis between portal areas can be measured, thereby providing an indicator of drug efficacy based on the pattern of fibrosis progression / regression in the patient.

[0052] The number and density of gate regions for each patient, each slide, each sample, each tissue region, or each repeat biopsy within a slide can be used as quality control indicators to assess the eligibility of samples for pathological evaluation.

[0053] Spatial indices based on the type, location, area, boundaries, or combinations thereof, or patterns of portal regions or other liver microfractions, can be used for correlation analysis with outcome or histopathological scores. Spatial indices generated by this method are preferably used for fully automated digital scoring of liver diseases.

[0054] The results of segmentation that identifies the component features of the gate region can be used to generate indicators related to the component features, as well as scores related to symptoms, diseases, or functional impairments associated with those component features.

[0055] According to one aspect of the invention, a system is provided comprising a computer program that can be directly loaded into a computer's memory, for operating the method according to any one of the preceding claims. The system may comprise a computer with the computer program loaded into memory.

[0056] The invention will now be described by way of example only, with reference to the accompanying drawings, in which:

[0057] Brief description of the attached figures

[0058] Figure 1 The background and foreground of liver tissue on histological sections are shown;

[0059] Figure 2 The portal region of liver tissue is shown on a histological section of the liver.

[0060] Figure 3 The gate regions of adjacent colocalized anatomical micro-components are shown;

[0061] Figure 4 A schematic diagram of the gate region and its associated micro-components is shown;

[0062] Figure 5 A schematic diagram illustrating the patterns of microcomponents in liver tissue is shown; and

[0063] Figure 6 A flowchart illustrating the methods for interpreting histological anatomy is shown.

[0064] Detailed description of the invention

[0065] Figure 1 The histology of liver tissue on histological section 700 is shown. Everything within the tissue outline is the tissue foreground 17. There is nothing visible outside the tissue itself because the tissue background 18 is simply the glass of the histological section.

[0066] Figure 2 and Figure 3A portion of a liver histological section is shown, in which liver tissue has been stained to highlight the various microfractions. The outlines of portal regions 101 and 102 (also known as portal triads) are visible. Typically, the various microfractions contained within portal regions 101 and 102 include at least one hepatic artery 41, 42, and 45; at least one portal vein 51, 52, and 55; and at least one bile duct 201, 202, 203, 204, and 208. Figure 2 The specific portal region 101 shown includes two hepatic arteries 41 and 42, two portal veins 51 and 52, and one bile duct 201. Connective tissue 31 also exists within the portal region 101, surrounding the various micro-components: the hepatic arteries, portal veins, and bile duct, and holding the portal region 101 together.

[0067] Connective tissue 31, 32, 33, 34, 35, 36, 37, 38, 39 in Figures 1 to 5 Visible within the liver is the connective tissue, the matrix surrounding the liver capsule (Glisson), which is composed of connective tissue and blood vessels. The connective tissue forms a reticular structure within the liver, providing integrity for hepatocytes and hepatic sinusoids. Portal canal aggregates 101-120, containing micro-components of the bile ducts, portal vein, and hepatic artery, are located within the connective tissue. The boundaries of this connective tissue define the boundaries of the portal canal.

[0068] Interfaces 161 and 162 exist along the defining boundary between the liver tissue inside the portal regions 103 and 104 and the liver tissue 35 outside the portal regions. Figure 2 and Figure 3 In liver histology, the interfaces 161 and 162 between the portal regions 103 and 104 and the surrounding external liver tissue are difficult to see. (Schematic diagram) Figure 4 The interface 16 surrounding the portal regions 103 and 104 is shown by dashed lines to indicate a gradual and difficult-to-detect change between the liver tissue inside the portal regions 103 and 104 and the liver tissue 35 outside.

[0069] Bile duct 201-208 in Figure 2 and Figure 3 It can be seen in histology, and in Figure 4 The diagram illustrates the bile ducts, which are pathways in the liver that facilitate the transport of bile. The bile ducts are micro-fragments of the portal tract 101-120. The bile ducts are characterized by their lining with specialized and visually distinctive epithelial cells called bile duct epithelial cells.

[0070] Hepatic artery 41-45 Figure 2 and Figure 3 It is visible in histology, and Figure 4The hepatic arteries 41-45 are schematically shown in the diagram. They are characterized by an outer layer composed of collagen fibers, a middle layer of smooth muscle, and an inner layer of endothelium lined with endothelial cells.

[0071] Portal vein 51-55 Figure 2 and Figure 3 It can be seen in histology, and in Figure 4 The diagram illustrates this. The portal veins 51-55 originate posterior to the neck of the pancreas, supplying the liver with oxygen-poor but nutrient-rich blood. Unlike the systemic veins, the portal veins 5 and their draining veins carry blood to the liver before it reaches the heart. The portal vein is characterized by its broad lumen lined with a single layer of flattened endothelial cells, numerous portal vein branches, and thin walls of fibrous connective tissue.

[0072] Anatomical micro-components in Figure 1 , Figure 2 and Figure 3 They are visible in liver histological sections. Figure 4 and Figure 5 The schematic diagram of liver histology is shown. Anatomical micro-components refer to the constituent elements that make up the tissue foreground 17, as well as aggregates of micro-components. Anatomical micro-components include, but are not limited to: portal areas 101-120, bile ducts 201-207, connective tissue 31-37, hepatic artery 41-45, portal vein 51-55, central vein 61-64, acini, hepatic lobules 81-84, hepatic regions, acinar regions 10, 11, 12, fibrotic bridging 151, 152, collagen deposition, interface regions 161, 162, and co-located anatomical micro-components, such as fatty degeneration 91, hepatocytes 92, inflammatory cells 93, and balloon cells 94.

[0073] Collagen deposition is an excessive or abnormal accumulation of collagen and other proteins in the extracellular matrix caused by liver fibrosis. Figure 4 The schematic diagram illustrates a fibrotic bridging 151 connecting two portal regions 103 and 104. Collagen deposition is a hallmark of liver fibrosis. Collagen deposition is used to detect and identify fibrotic bridging 151, and to detect and identify... Figure 1 The foreground of the histological section shown is the portal region 101-120 in the liver tissue (17).

[0074] Figure 5 A schematic diagram of liver histology is shown. This diagram is intended to visually explain... Figure 1 , Figure 2 and Figure 3 The liver histological sections reveal various micro-components within the liver tissue and the complex arrangement of their aggregates. From Figure 5 The complex patterns are visible in the tissue and are easier to see than in liver histological sections.

[0075] Figure 5 One pattern in the liver lobules is the hexagonal arrangement of lobules 81-84. Lobules 81-84 are micro-components of liver histology. At each of the six peripheral angles of each roughly hexagonal lobule 81-84, there is approximately a portal area 105-120°. The hexagonal arrangement of lobules 81-84 presents a hexagonal honeycomb pattern. At the center of each hexagon is approximately a central vein 61-64.

[0076] It should be understood that the complex arrangement of portal zones 105-120 in liver tissue in histological sections is not a precise hexagon. Lobules 81-84 also do not have a precise hexagonal shape. Portal zones 105-120 in liver microanatomy are distributed in a meandering biological pattern. Furthermore, in liver tissue, the central veins 61-64 are only approximately located at the center of lobules 81-84. In histological sections of liver tissue, for personnel or machines trained to recognize honeycomb patterns, portal zones 105-120 appear to be roughly at the corners of a hexagon. Figure 5 The diagrams in the diagrams are helpful for training in locating portal tract 1, central vein 61-64, and other micro-components in histological sections. The approximate hexagonal pattern also helps in confirmation; the discovery of aggregates of micro-components is likely to be the discovery of portal tract 105-120, as the aggregates are located approximately at the corners of the hexagons that are readily apparent from other discovered portal tract 105-120.

[0077] exist Figure 5 Another pattern that can be seen is a nested rhombus pattern of three acinar partitions 10, 11, and 12. The largest of the three is the third acinar partition 12. The second largest is the second partition 11, which is nested within the third acinar partition 12. The smallest is the first partition 10, which is nested within the second partition 11. All three rhombuses have a first pair of opposite corners 181 and 182 and a second pair of opposite corners. All three rhombuses share the first pair of opposite corners 181 and 182. Each of the two opposite corners 181 and 182 shared by all three rhombuses has a gate tube region.

[0078] The three rhombuses in the three acinar regions 10, 11, and 12 do not share the second pair of diagonals because they are nested. The largest of the three rhombuses is the third acinar region 12, in which central veins 61 and 63 are located at each of the second pair of diagonals 171 and 172. Central veins 61 and 63 are located in adjacent hepatic lobes 82 and 84. The second pair of diagonals 173 and 174 of the second acinar region 11 lie on an imaginary line passing through the second pair of diagonals 171 and 172 of the third acinar region 12. That is, this imaginary line passes through central veins 61 and 63. The second pair of diagonals 175 and 176 of the first acinar region 10 lie on this imaginary line, such that the three acinar regions 10, 11, and 12 are nested. The second pair of diagonals 173 and 174 of the second acinar partition 11 are distributed between the corresponding corner points of the second pair of diagonals 171, 172, 175 and 176 of the first acinar partition 10 and the third acinar partition 12, so that the three acinar partitions 10, 11 and 12 are nested.

[0079] It should be understood that the complex arrangement of adjacent central veins 6 and 61, adjacent portal canal areas 181 and 182, and the second pair of diagonals 171, 172, 173, 174, 175, and 176 in liver histological sections is not a precise rhombus. These angles are curved rather than sharp. The outline of each rhombus is meandering rather than straight. The nested rhomboid pattern is not precisely concentric. Figure 5 The diagram illustrating nested diamond patterns is helpful for training in the search of portal regions 181, 182, central veins 61, 63, hepatic lobes 82, 84, acinar regions 10, 11, 12, and other micro-components in liver tissue within histological sections. These patterns are designed to assist personnel or machines in detecting, identifying, or recognizing specific micro-components within the patterns. This assistance is useful because simple segmentation of micro-components or aggregates of micro-components in liver tissue of liver technical sections often does not provide sufficient information for identification or recognition with a sufficiently accurate probability.

[0080] The acinar subdivision furthest from the central veins 61 and 63 (called acinar subdivision 10) receives the most oxygenated blood and the most nutrients. The acinar subdivision closest to the central veins 61 and 63 (called acinar subdivision 32) receives the least oxygenated blood and the least nutrients. Acinar subdivision 21, located between acinar subdivision 10 and acinar subdivision 32, receives a moderate amount of oxygenated blood and a moderate amount of nutrients.

[0081] Figure 6 A flowchart of a method 600 for histological anatomical interpretation is shown. By way of exemplary examples, the portal regions 101-120 in a histological image of a histological section 700 (e.g., H&E, Masson trichrome, Sirius red, or any other staining) are identified and characterized according to the following scheme:

[0082] 1. Create or use semantic segmentation, instance segmentation, or other segmentation models for each liver tissue:

[0083] 1.1. Segmentation of foreground tissue 604;

[0084] 1.2. Segmentation of connective tissue 606;

[0085] And specific micro-component 602, especially gate region micro-component 608;

[0086] 1.3. Division of the bile duct at 610;

[0087] 1.4. Segmentation of hepatic artery 612; and

[0088] 1.5. Division of the portal vein 614.

[0089] 2. The target objects of these different segmentation problems have known or anticipated spatial relationships. Analysis is performed according to histological characterization methods and systems to produce segmentations of complex tissue aggregates of liver micro-components that are portal regions. These segmentations (all or several, or one or more subsets thereof) are spatially correlated to determine the confidence or probability that a region of connective tissue 31-39 is a portal region 101-120. For example, in one embodiment, a connective tissue region is considered a portal region only if it contains a prescribed or preselected number of bile ducts, portal veins, or hepatic arteries (individually or in total).

[0090] 3. Semantic segmentation of micro-components or their aggregates is processed into one or more predicted gate regions. Candidate 618 of gate region examples are identified or selected from the intersection of tissue foreground 604 and connective tissue 606. The term "gate region" (PT) 101-120 refers to the region of segmented connective tissue (CT) 31-36 that intersects with one or more of the following examples:

[0091] a) Segmented bile ducts (BD) 201-208.

[0092] b) Segmented hepatic artery (HA) 41-45.

[0093] c) Divided portal vein (PV) 51-55.

[0094] 4. When a micro-component or aggregate {a|b|c} intersects with CT but is partially located outside the connective tissue segmentation region (i.e. partially inside 624), the boundary of the gate region is the union of the connective tissue and the micro-component or aggregate {a|b|c} 626.

[0095] 5. The intersection of the target and connective tissue is calculated using a distance threshold associated with the boundary of the connective tissue. For example, if a portal vein is detected near a region of connective tissue (e.g., within 5, 10, or 20 micrometers) but does not intersect with the original connective tissue, the portal vein is still considered part of the portal canal region, and the final portal canal region is the union of the connective tissue and the portal vein. This verifies whether the 620 micro-component target intersects with the candidate PT boundary or is within a distance threshold of the candidate PT boundary.

[0096] 6. Count the components that meet the criteria for candidate gate regions, including those completely within the candidate gate region 628 and those that are the union of the candidate gate regions 626. Assign a reliability score 632 to the candidate gate differentiation based on the number of different features associated with the boundary connective tissue.

[0097] a) For ≥1 bile duct, +1

[0098] b) For ≥1 portal vein, +1

[0099] c) For ≥1 hepatic artery, +1

[0100] 7. Therefore, based on which components of the portal triad (portal vein, bile duct, or hepatic artery) are identified as present in the region, reliability is assigned to each candidate portal region segment:

[0101] 0 = an element in a gateless triplet.

[0102] 1 = Single element, namely bile duct, portal vein or hepatic artery.

[0103] 2 = Any two of the three components of the gate tube triplet.

[0104] 3 = All components of the gate tube triplet.

[0105] A higher score corresponds to a higher prediction probability. This confidence score provides an interpretable basis for defining image regions as gate regions and can be used to filter candidate gate regions based on the number of associated features. This can be regarded as a proxy indicator of the probability that candidate micro-component aggregates are actually gate regions.

[0106] A pre-selected confidence score is used to ensure that the probability of a candidate gate region being real exceeds a minimum threshold 634. If the confidence score of a candidate gate region is lower than this minimum, the aggregate of elements and the connective tissue in which they reside are determined to be in connective tissue outside the gate region 636. If the confidence score of a candidate gate region is lower than this minimum, a gate region is determined to be detected and identified, and it includes the aggregate of elements and connective tissue of the candidate gate region 636.

[0107] Information about the identified portal areas (including their location in the tissue foreground) is combined with information determined from the tissue foreground (e.g., segmentation of the central vein in the tissue foreground) 638. This enables the location of hepatic lobular regions (e.g., acinar subdivisions) 640.

[0108] As previously explained, due to the presence of patterns of portal canals and central veins indicating possible locations of portal canals within the liver lobule regions, these patterns can be fed back to step 618 for identifying candidate portal canal examples. These patterns can then be fed back to step 620, for example, to check whether the location of a micro-component or its aggregate intersects within a certain distance or angular range of a location in one of the patterns. This leads to improvements in histological characterization methods as the analysis of histological sections progresses, resulting in faster and more accurate detection of portal canals, central veins, and liver lobule regions, including but not limited to acinar partitions of tissue on sections.

[0109] Geometric analysis 636 is performed on the detected and identified portal region 642. Auxiliary liver regions 644 can be detected and identified, such as liver regions with steatosis 91, hepatocytes 92, inflammatory cells 93, ballooning cells 94, and fibrotic bridging 15. Geometric analysis 636 can be combined with connective tissue identified as being outside the portal region by confidence analysis 636 or within intersection or distance thresholds 622 to find auxiliary regions outside the portal region but associated with it.

[0110] The present invention has been described by way of example only. Therefore, the foregoing is to be regarded only as an illustration of the principles of the invention. Furthermore, since it is easy for those skilled in the art to make numerous modifications and alterations, the invention is not intended to be limited to the exact constructions and operations shown and described, and accordingly, all suitable modifications and equivalents falling within the scope of the claims can be employed.

Claims

1. A method for determining liver histological parameters, comprising: Segmentation of histological micro-components from biological samples using liver histological images; The micro-components that are within a pre-selected proximity to each other are spatially associated as aggregates of the micro-components, the aggregates of the micro-components defining histological components or indicators of disease severity. as well as Each of the micro-components within a pre-selected proximity to each other is associated with the bile duct, portal vein, and hepatic artery, such that the aggregates of the components have a confidence level of over a threshold as a portal region.

2. The method of claim 1, wherein the presence or absence of the micro-components within a pre-selected proximity affects an indicator or confidence score of the probability that the aggregates of the components are "true" gate regions.

3. The method according to claim 1 or 2, wherein the use of the micro-components, the aggregates, and the spatial domains is isolated from other spatial features of the biological sample.

4. The method according to any of the preceding claims, wherein the aggregates identified as gate areas are quantified to assess sample eligibility.

5. The method according to any of the preceding claims, wherein the location of inflammatory cells, balloon cells, fibrosis, fat globule features, or connective tissue is integrated with the liver lobules and the aggregates as portal zones to provide an indicator of spatial differentiation of the number and density of inflammatory cells within and outside the portal zones, or to identify bridging fibrosis between the portal zones.

6. The method according to any of the preceding claims, wherein a segmentation model is used to identify the micro-components of the gate region or as the aggregates of the gate region.

7. The method of claim 6, wherein the micro-components or aggregates are segmented using a neural network, a convolutional neural network, a Transformer model, including a visual Transformer, or a combination of multiple such networks trained for the purpose.

8. The method of claim 6, wherein the micro-components or aggregates are segmented using a neural network containing a segmentation base model, or an autoencoder or Transformer capable of learning and generating embedding vectors of arbitrary image features.

9. The method of claim 6, wherein the micro-components or aggregates are segmented using non-neural network-based image processing via thresholding, pixel classification, filtering, or feature detection.

10. The method according to any one of claims 6 to 9, wherein micro-components or aggregates are roughly segmented by using patch-based classification, wherein smaller patches are extracted from a larger image and classified according to whether the patches contain the presence of each micro-component.

11. The method according to any of the preceding claims, wherein the segmentation of the micro-components or aggregates is identified as a summary set of bounding boxes, centroids, or other coordinates, and these summary descriptors are spatially correlated.

12. The method according to any of the preceding claims, wherein the micro-component or aggregate is represented as a position-based probability of the presence of the component feature at a corresponding location in a probability graph, and spatial association of the micro-component or aggregate is performed by a geometry derived from thresholding the probability graph at a certain value.

13. The method according to any of the preceding claims, wherein the micro-components or aggregates are identified by segmenting the cell types to which they consist using a cell segmentation or detection model trained to predict cell types.

14. The method of claim 13, wherein the micro-components or portions of the aggregates of connective tissue are identified as connective tissue cells within proximity to each other.

15. The method according to any of the preceding claims, wherein the aggregates as gate regions and the hepatic lobules of the liver are delineated by a combination of micro-components of the aggregates and foreground features of the tissue via logical predicates including union, intersection, or difference.

16. The method according to any of the preceding claims, wherein the interface region is delineated between the aggregate, which is a gate region, and the hepatic lobules of the liver through constructive and destructive geometric operations including features of buffering, expansion, or corrosion.

17. The method according to any of the preceding claims, wherein the acinar regions to be analyzed in a liver histological image are delineated by subdividing the aggregates into portal regions and lobular regions of the liver.

18. The method of claim 17, wherein a midline is found between two gate regions and a leaflet region closest to one of the two gate regions, and further subdivided by radial distance from one boundary of the gate region to the other boundary.

19. The method according to any one of the preceding claims, wherein the method uses a gate region model to distinguish between the gate region, leaflet region, and interface region by generating descriptive statistics.

20. The method according to any one of the preceding claims, wherein the method expresses the distribution of portal inflammation, lobular inflammation or interface inflammation according to at least one of the following: the number of inflammatory foci, the average area of ​​the foci, the density of inflammatory cells in the portal area, the density of inflammatory cells in the interface or the density of inflammatory cells in the lobular area.

21. The method according to any one of the preceding claims, wherein the method identifies the spatial location of neutrophils, eosinophils, lymphocytes (including T cells and B cells), plasma cells, macrophages (including Kupffer cells) or other inflammatory cell types relative to the portal area, lobular region and other liver partitions.

22. The method according to any one of the preceding claims, wherein the method identifies the location and abundance of clusters of cooperating cells relative to a gate region, leaflet region, or interface region as defined by statistical frequency, nearest neighbor, "bag of words," or other techniques.

23. The method according to any one claim, wherein the method combines the gate regions with the component characteristics of the connective tissue to identify bridging fibrosis between the gate regions.

24. The method according to any one of the preceding claims, wherein the method measures the component characteristics of the connective tissue relative to the shape, length and other morphological characteristics of the gate region to assess the formation of collagen bridging.

25. The method according to any one claim, wherein the biological sample in the liver histological image is a longitudinal sample to assess changes in the portal region, lobular region, interface, or connective tissue over time, including assessing the reduction or increase of inflammation, fibrosis, and scarring.

26. The method according to any of the preceding claims, wherein the length and width of the bridging fibrosis between the gate regions are measured, thereby providing an indicator of drug efficacy based on the pattern of fibrosis progression / regression in the patient.

27. The method according to any of the preceding claims, wherein the number and density of gate regions within the slice for each patient, each slide, each sample, each tissue region, or each repeat biopsy are used as quality control indicators to assess the eligibility of samples for pathological evaluation.

28. The method according to any of the preceding claims, wherein the method uses spatial indicators based on the location / area / boundaries of the portal region and other liver subcompartments to perform correlation analysis with the outcome or histopathological score.

29. The method of claim 28, wherein the spatial index generated by the method is used for a fully automated digital scoring of liver diseases.

30. The method according to any of the preceding claims, wherein the results of segmentation of identifying component features of the gate region are used to generate an index associated with said component features, and a score associated with the symptoms, diseases, or functional impairments of said component features is generated.

31. A system comprising a computer program that can be directly loaded into a computer's memory for operating the method according to any one of the preceding claims.

32. The system of claim 30, comprising a computer having the computer program loaded into the memory.

Images