Predicting the presence of histopathological abnormalities

Abstract

The present invention relates to the application of toxicogenomics methods to the detection and / or prediction of histopathological abnormalities in human or animal subjects based on clinical pathology data. It has been observed that reliable results can be obtained even in the absence of image data. A computer-implemented method for predicting the presence of a histopathological abnormality in an organ of a human or animal subject based on clinical pathology data comprises receiving clinical pathology data obtained from the human or animal subject, applying an analytical model to the clinical pathology data, the analytical model being configured to output a result indicative of the likelihood of the presence of the histopathological abnormality in the organ of the human or animal subject, and outputting the result.

Description

[Technical Field]

[0001] The present invention relates to a computer-implemented method for identifying histopathological abnormalities in human or animal subjects, and corresponding training methods and systems are also provided. [Background technology]

[0002] Drug development involves the safety evaluation of test compounds in animals to determine their safety in humans. Preclinical toxicity studies consist of lifetime, laboratory, molecular, and postmortem evaluations in animals such as rodents, dogs, and non-human primates. Toxicogenomics studies are toxicity studies in which gene expression in specific organs is correlated with standard toxicological endpoints such as clinical pathology and histopathology, as in Uehara et al. (2010).

[0003] As with human pathologies, there are an increasing number of applications of digital and computational technologies in toxicological pathologies, see Abels et al. (2019), Turner et al. (2020), Turner et al. (2021), and Mehrvar et al. (2021).

[0004] These new technologies have a significant impact on the timeline of histopathological evaluation. Furthermore, they help improve the quality of histopathological data by: improving the quality, reproducibility, and rigor of histopathological data; uniformly setting thresholds for morphological changes in control tissues; and shortening timelines. Computational pathology is not limited to the detection of lesions or morphological patterns of lesions, but also involves the integration, complex analysis, and interpretation of a wide range of assays for the diagnosis, treatment, and prognosis of disease. [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Uehara, T., Ono, A., Maruyama, T., Kato, I., Yamada, H., Ohno, Y., and Urushidani, T. (2010), "Japan's Toxicogenomics Project: Applications of Toxicogenomics." Mol. Nutr. Food Res., 54:218-227. https: / / doi.org / 10.1002 / mnfr.200900169 [Non-patent document 2] Abels, Esther et al. (2019) "Computational Pathology Definition, Best Practices, and Recommendations for Regulatory Guidance: A White Paper from the Digital Pathology Society." Journal of Pathology 249.3 (2019): 286-294. [Non-patent document 3] Turner, Oliver C. et al., "Toxicologic Pathology Digital Pathology and Image Analysis Special Interest Group Paper*: Comments on the Application of Artificial Intelligence and Machine Learning to Digital Toxicologic Pathology," Toxicologic Pathology 48.2 (2020): 277-294 [Non-patent document 4] Turner, Oliver C. et al. "Mini Review: The Last Mile - Opportunities and Challenges of Machine Learning in Digital Toxicologic Pathology." Toxicologic Pathology 49.4 (2021): 714-719 [Non-Patent Document 5] Mehrvar S, Himmel LE, Babburi P, Goldberg AL, Guffroy M, Janardhan K, Krempley AL, Bawa B. "Deep learning approaches and applications in toxicological histopathology: current status and future perspectives." JJ Pathol Inform [serial online] 2021 [cited 16 November 2021];12:42. Available from: https: / / www.jpathinformatics.org / text.asp?2021 / 12 / 1 / 42 / 329733 Summary of the Invention

[0006] The present invention relates to the application of toxicogenomics methods to the detection and / or prediction of histopathological abnormalities in human or animal subjects based on clinical pathology data. Importantly, it has been found that reliable predictions can be made in the absence of image data. Accordingly, a first aspect of the present invention provides a computer-implemented method for predicting the presence of a histopathological abnormality in an organ of a human or animal subject based on clinical pathology data, the computer-implemented method comprising receiving clinical pathology data obtained from the human or animal subject, applying an analytical model to the clinical pathology data, the analytical model being configured to output a result indicative of the likelihood of the presence of the histopathological abnormality in the organ of the human or animal subject, and outputting the result.

[0007] The term "predicting" may refer to determining at least one numerical or categorical value that indicates the presence of a histopathological abnormality.

[0008] The term "subject" as used herein typically relates to mammals, but can also refer to other classes of animals. The subject may be suffering from a disease or may be suspected of suffering from a disease, i.e., the subject may already exhibit some or all of the negative symptoms associated with the disease. In the present application, the organ of the human or animal subject is preferably the liver. However, the computer-implemented method is equally applicable to other organs, such as the kidney. There are various types of histopathological abnormalities, the presence of which can be detected using the computer-implemented method of the first aspect of the present invention. However, in preferred cases, the histopathological abnormality is a lesion.

[0009] As mentioned above, it has been found that reliable predictions can be made in the absence of image data. Thus, by using the computer-implemented method of the present invention, it is possible to reliably predict the presence of histopathological abnormalities in organs of human or animal subjects without the need for invasive or expensive procedures required to obtain histopathological slide images. For completeness, it is prudent to state that in preferred embodiments of the present invention, clinical pathology data does not include image data. As used herein, "image data" refers to data, such as electronic data, representing an image of a region of an organ of interest. The image may be a photograph of a histopathological slide or may be obtained using a range of well-known medical imaging techniques.

[0010] Alternatively, in a preferred embodiment of the present invention, the analytical model may be performed solely on clinical pathology data. In the context of the present application, "clinical pathology data" is used to refer to data that can be obtained non-invasively and generally relates to the presence and amount of one or more analytes in the bodily fluids of a human or animal subject. The bodily fluids may include samples of the subject's tissues / organs and / or samples of products produced by the subject's tissues / organs. The products produced by the subject's tissues / organs may be, for example, secretions (e.g., glandular secretions, milk, colostrum, tears, saliva, sweat, earwax, mucus), sputum, semen, vaginal / cervical fluid, blood (plasma, serum), cerebrospinal fluid (CSF), excretions, feces or urine, skin or hair products.

[0011] Clinical pathology data may include measurements of the concentration of one or more analytes in a bodily fluid of a human or animal subject. The analytes may comprise one or more biomarkers. As used herein, the term "biomarker" refers to a biological molecule found in blood, other bodily fluids, or tissues that is indicative of a normal or abnormal process, or a condition or disease. In some cases, clinical pathology data may include measurements of the concentration of one or more analytes in one or more bodily fluids. For example, a first subset of clinical pathology data may include measurements of the concentration of one or more analytes in a first bodily fluid, and a second subset of clinical pathology data may include measurements of the concentration of one or more analytes in a second bodily fluid. This can, of course, be generalized to multiple bodily fluids in general.

[0012] The one or more biomarkers may include one or more of a liver injury biomarker, a muscle injury biomarker, and a kidney injury biomarker. As used herein, the term "injury biomarker" refers to a biomarker that is present in higher concentrations when an abnormality or other injury is present in an organ or tissue of a subject.

[0013] The analyte may also include one or more electrolytes.

[0014] The clinical pathology data may include a ratio of the concentration of a first analyte to the concentration of a second analyte (or vice versa). The clinical pathology may include a plurality of such ratios. In particular, the clinical pathology data may include a ratio of the concentration of albumin in the bodily fluid of the human or animal subject to the concentration of globulin in the bodily fluid of the human or animal subject. Alternatively, for completeness, the clinical pathology data may include a ratio of the concentration of globulin in the bodily fluid of the human or animal subject to the concentration of albumin in the bodily fluid of the human or animal subject. This ratio is typically used to help clinicians identify the cause of changes in protein levels in a user's bodily fluid.

[0015] Liver injury biomarkers may include one or more of bilirubin; aspartate aminotransferase (which may also be called "glutamic oxaloacetic transaminase"); gamma glutamyltransferase (which may also be called "gamma-glutamyltranspeptidase"); alanine aminotransferase (which may also be called "glutamic pyruvic transaminase"); and lactate dehydrogenase. In preferred cases, liver injury biomarkers may include all of bilirubin; aspartate aminotransferase; gamma glutamyltransferase; alanine aminotransferase; and lactate dehydrogenase. Liver injury biomarkers may further include one or more of the following: albumin; alkaline phosphatase; cholesterol; globulin; glucose; protein; and triglycerides.

[0016] If preferred, the electrolyte may include potassium. The electrolyte may further include one or more of the following: calcium; chloride; phosphate; and sodium.

[0017] In preferred cases, muscle damage biomarkers may include creatine kinase, where high levels of creatine kinase in the blood generally indicate recent muscle damage.

[0018] Kidney damage biomarkers may include one or more of the following: creatinine and urea nitrogen. These biomarkers are commonly used to assess kidney function.

[0019] In a particularly preferred embodiment of the first aspect of the present invention, the clinical pathology data comprises the ratio of the concentration of albumin in the body fluid of the human or animal subject to the concentration of globulin in the body fluid of the human or animal subject, and the concentrations of each of the following in the body fluid of the human or animal subject: bilirubin; aspartate aminotransferase; gamma glutamate transferase; alanine aminotransferase; lactate dehydrogenase; potassium; and creatine kinase.

[0020] Having described the nature of clinical pathology data in detail, more information regarding analytical models will now be provided. When applied to clinical pathology data, analytical models are configured to output results indicative of the probable presence of histopathological abnormalities in organs of a human or animal subject. As used herein, the term "analytical model" may refer to a mathematical model configured to predict at least one target variable with respect to at least one state variable. The term "target variable" may refer to a clinical value to be predicted. The target variable value to be predicted may depend on the disease or condition whose presence or state is to be predicted. The target variable may be numerical or categorical. For example, the target variable may be categorical and may be "positive" if the disease is present or "negative" if the disease is absent. As used herein, the term "state variable" may refer to an input variable that may be input into a predictive model, such as data derived from a medical examination and / or self-examination by the subject. The state variable may be determined in at least one active test and / or at least one passive monitoring.

[0021] The target variable may be a numerical value, such as at least one value and / or scale value, in which case the state variable may include clinical pathology data and the target variable may include an indication of the likelihood of the presence of a histopathological abnormality in an organ of a human or animal subject.

[0022] The analytical model may be a regression model or a classification model. In the context of this application, the term "regression model" may be used to refer to an analytical model whose output is a numerical value within a range. For example, the output of such a regression model in this case may be a numerical value corresponding to the likelihood or probability of the presence of a histopathological abnormality in an organ of a human or animal subject. In the context of this application, the term "classification model" may be used to refer to an analytical model whose output is a binary classification or score indicating the presence or absence of a histopathological abnormality in an organ of a human or animal subject.

[0023] Specifically, the analytical model may be a machine learning model trained to output a result indicating the possible presence of histopathological abnormalities in an organ of a human or animal subject based on an input including clinical pathology data obtained from the human or animal subject. The machine learning model may be a regression model or a classification model, as defined above. The machine learning model is preferably trained using supervised learning. Thus, in a preferred embodiment, the machine learning model is a random forest model or a gradient boosting model. Herein, when a machine learning model is referred to as a "random forest model," it means that the machine learning model has been trained using a random forest algorithm. Similarly, when a machine learning model is referred to as a "gradient boosting model," it means that the machine learning model has been trained using a gradient boosting algorithm.

[0024] The Random Forest algorithm is a supervised learning algorithm that combines the outputs of multiple decision trees to arrive at a single result. When the Random Forest algorithm is used for regression, the output may include the average or mean prediction of the individual decision trees. When the Random Forest algorithm is used for classification, the output may include the class selected by the most individual decision trees. Breiman, L. Random Forests, Machine Learning 45, 5-32 (2001), https: / / doi.org / 10.1023 / A:1010933404324, Ooka T, Johno H, Nakamoto K, et al. A Random Forest Approach for Determining Risk Prediction and Predictors of Type 2 Diabetes: Large-Scale Health Checkup Data from Japan BMJ Nutrition, Prevention & Health 2021;bmjnph-2020-000200.doi:10.1136 / bmjnph-2020-000200, and "Random Forest Classification Algorithms for Healthcare Industry Data" Christodoulos Vlachas, Lazaros Damianos, Nikolaos Gousetis, Ioannis Mouratidis, Dimitrios Kelepouris, Konstantinos-Filippos Kollias, Nikolaos Asimopoulos and George F Fragulis; SHS Web Conf., 139 (2022) 03008, DOI: https: / / doi.org / 10.1051 / shsconf / 202213903008 may be used. Other examples of random forest algorithms are suitable as well.

[0025] The gradient boosting algorithm is a supervised learning algorithm that generates a predictive model in the form of an ensemble of weak predictive models, such as decision trees. Gradient boosted tree models are constructed incrementally, similar to other boosting methods, but generalize other methods by allowing the optimization of any differentiable loss function. Examples of gradient boosting algorithms that may be used include XGBoost.

[0026] The analytical model is preferably configured to output a histopathological score indicative of the likelihood of the presence of a histopathological abnormality in an organ of a human or animal subject. This is true whether the analytical model is a regression model or a classification model. In preferred cases, the histopathological score is a binary score (e.g., "1" or "0", although it will be appreciated that any binary score can be used). Here, a "binary score" is a score that can take only two values. Preferably, one value corresponds to a prediction of the presence of a histopathological abnormality, and the other value corresponds to a prediction of the absence of a histopathological abnormality.

[0027] A second aspect of the present invention provides a computer-implemented method for generating a machine learning model configured to output a result indicating the likelihood of the presence of a histopathological abnormality in an organ of a human or animal subject, the computer-implemented method comprising: receiving, for each of a plurality of human or animal subjects, training data including clinical pathology data and a histopathological score indicating whether a histopathological abnormality is present in the organ of the human or animal subject; and training a machine learning model using the received training data. Except where clearly incompatible, optional features described above with respect to the first aspect of the present invention apply equally well to the second aspect of the present invention. In particular, training the machine learning algorithm may include using a random forest algorithm or a gradient boosting algorithm. Preferably, the analytical model of the first aspect of the present invention is a machine learning model generated according to the computer-implemented method of the second aspect of the present invention.

[0028] In some cases, the training data may include multiple subsets of training data, each set originating from a different source. In these cases, it may be desirable to standardize the training data, thereby reducing variability between the subsets of training data from different sources. Data standardization may also reduce variability within a given subset of training data. Therefore, prior to training the machine learning model, the computer-implemented method may further include normalizing the received training data. This may be done using various normalization techniques, including log-normalization, batch normalization, z-score normalization, and the use of a location scale model.

[0029] It is known that the effectiveness of a machine learning model can depend on the quality of training data used to train the machine learning model. Thus, a computer-implemented method for generating a machine learning model configured to output a result indicating the likelihood of a histopathological abnormality in an organ of a human or animal subject may include receiving, for each of a plurality of human or animal subjects, training data including clinical pathology data and a histopathological score indicating whether a histopathological abnormality is present in the organ of the human or animal subject, training a first machine learning model using the training data in a first manner, training a second machine learning model using the training data in a second manner, and selecting either the first trained machine learning model or the second trained machine learning model based on a performance metric.

[0030] In some cases, the first and second aspects may refer to training each machine learning model using different subsets of training data. Specifically, the training data may include a first subset of training data and a second subset of training data. The first machine learning model may then be trained using the first subset of training data, and the second machine learning model may be trained using the second subset of training data. The first subset of training data and the second subset of training data may not overlap, i.e., no individual subject may be represented in both the first subset of training data and the second subset of training data. Alternatively, the first subset of training data may partially overlap with the second subset of training data. However, the first subset of training data should not be identical to the second subset of training data.

[0031] In other cases, the first and second aspects may refer to training each machine learning model using different training algorithms. Specifically, the first machine learning model may be trained using a first training algorithm, and the second machine learning model may be trained using a second training algorithm. For example, the first training algorithm may be a random forest algorithm, and the second training algorithm may be a gradient boosting algorithm. Alternatively, both the first and second training algorithms may be random forest training algorithms (albeit different), or both the first and second training algorithms may be gradient boosting algorithms (again, different).

[0032] This may be further generalized, i.e., the computer implementation may include training each of the multiple machine learning models in a respective manner. Then, the final step may include selecting one of the multiple trained machine learning models based on a performance metric. "Each respective manner" may refer, for example, to a combination of a particular subset of training data and a training algorithm. In this way, when there are various training data available, it is possible to identify the subset of training data that results in the best performance. Similarly, when there are various training algorithms available for a given machine learning model, this allows a user to select the best-performing training algorithm.

[0033] We now discuss performance metrics. There are a variety of metrics available for evaluating the performance of machine learning algorithms (https: / / scikit-learn.org / stable / auto_examples / miscellaneous / plot_roc_curve_visualization_api.html). In preferred cases, the performance metric is the area under the receiver operating characteristic curve (commonly shortened to AUC or AUC-ROC), which quantifies the ability of a classification model to perform classification accurately. Alternative performance metrics that may be used include F1 score, precision, recall, and classification accuracy.

[0034] To obtain performance metrics, it is often necessary to apply the trained machine learning model to test data. Preferably, there is no overlap between the test data and the training data used to train the model to avoid training bias. Therefore, the computer-implemented method may further include determining a first value of the performance metric for a first trained machine learning model, determining a second value of the performance metric for a second trained machine learning model, and selecting one of the first trained machine learning model or the second trained machine learning model associated with a better score. Herein, "better" refers to a more favorable score, i.e., a score indicating better performance. While this often results in a higher score, in some cases, a lower score may indicate better performance. In either case, determining the respective values of the performance metrics may include applying the trained machine learning model to the test data and determining the values of the performance metric based on the output of the machine learning model. Preferably, each trained machine learning model is applied to the same test data to ensure consistency across various tests.

[0035] The foregoing aspects of the invention relate to computer-implemented methods. A further aspect of the invention may provide a system for predicting the presence of histopathological abnormalities in an organ of a human or animal subject based on clinical pathology data, the system comprising a processor configured to perform the computer-implemented method of any preceding aspect of the invention. Optional features described in relation to the preceding aspects of the invention apply equally well to this system aspect of the invention, unless clearly inconsistent.

[0036] A further aspect of the present invention may provide a computer program comprising instructions which, when the program is executed by a computer or a processor thereof, cause the computer or a processor thereof to perform the steps of the computer-implemented method of the previous aspect of the invention. Optional features described in relation to the previous aspect of the invention apply equally well to this aspect of the invention, unless clearly inconsistent. A further aspect of the present invention provides a computer-readable medium having stored thereon the computer program of the previous invention.

[0037] The present invention includes combinations of the described embodiments and preferred features except where such combinations are expressly not permitted or explicitly avoided. [Brief explanation of the drawings]

[0038] Embodiments of the present invention will now be described with reference to the accompanying drawings.

[0039] [Figure 1] 1 is a schematic diagram of a system that may be used in an embodiment of the present invention. [Figure 2] 1 is a flowchart illustrating the generation of a machine learning model that can be used to predict the presence or absence of a histopathological abnormality. [Figure 3] 1 is a flowchart illustrating the use of a machine learning model to predict the presence or absence of a histopathological abnormality. DETAILED DESCRIPTION OF THE INVENTION

[0040] Aspects and embodiments of the present invention will now be discussed with reference to the accompanying figures. Further aspects and embodiments will be apparent to those skilled in the art. All documents mentioned in this text are incorporated herein by reference.

[0041] FIG. 1 shows an overall system 1 that can be used to perform a computer-implemented method according to the present invention. The system 1 includes a data acquisition unit 100, an analysis system 200, and a display module 300. The data acquisition unit 100, the analysis system 200, and the display module 300 are all interconnected via a network 400. The network 400 may be a wired network (such as a LAN or WAN) or a wireless network (such as a Wi-Fi network, the Internet, or a cellular network). In some cases, the data acquisition unit 100, the analysis system 200, and the display module 300 may be connected via multiple networks 400 (not shown). For example, the acquisition module 100 and the analysis system 200 may be connected via a first network, the data acquisition unit 100 and the display module 300 may be connected via a second network, and the analysis system 200 and the display module 300 may be connected via a third network. Other combinations are also contemplated. Alternatively, some subset of the data acquisition unit 100, analysis system 200, and display module 300 may be integrated with one another. For example, the data acquisition unit 100, analysis system 200, and display module 300 may all be incorporated into a single system, such as a smartphone, desktop computer, laptop computer, or tablet. In some cases, the data acquisition unit 100 and display module 300 may be client- or user-facing modules (i.e., they are accessible by end users or clients of the system), while the analysis system 200 may be located remotely, for example, on a server such as a back-end server. Alternatively, the analysis system 200 may be located on the cloud, such that processing is performed external to the user device on a server (or the like) with greater computing power. Various other configurations are envisioned.

[0042] In the context of the present application, the data acquisition unit 100 is a unit that may include hardware and software components adapted to acquire clinical pathology data from a human or animal subject. For example, a clinician or other scientist may use dedicated hardware to acquire samples of body fluids from the human or animal subject of interest and, for example, generate clinical pathology data using the acquired samples. Of course, in the context of the present invention, the clinical pathology data is preferably electronic data containing information regarding the concentrations of various analytes in the body fluids of the human or animal subject of interest, as outlined elsewhere in this patent application.

[0043] Analysis system 200 is where much of the analysis central to the present invention is performed. Analysis system 200 comprises processor 204 and memory 208. Processor 204 comprises a number of modules. In the context of this application, a module may be implemented in either hardware or software and is adapted or configured to perform a particular function. For example, a module may be used to refer to a physical component within a processor, or a module may refer to a section of code containing instructions that, when executed by processor 204, cause the processor to perform a function of interest. Specifically, processor 204 comprises a preprocessing module 2041, a training module 2042, a testing module 2044, a selection module 2046, and an analysis module 2048. The respective functions of each of these modules are described below.

[0044] Memory 206 of analysis system 200 may comprise both persistent and temporary memory. In the specific embodiment shown in Figure 1, memory 206 stores gradient boosting algorithm 2062, random forest algorithm 2064, training data 2066 (which includes two subsets 20662, 20664), and testing data 2068. Memory 206 also includes buffer 2070, which may store, for example, clinical pathology data 20702 while processing is taking place (described in more detail below).

[0045] Having described the structure of system 1, its operation will now be described with reference to Figures 2 and 3, which are high-level flowcharts illustrating a computer-implemented method for generating a machine learning model and a computer-implemented method for using a machine learning module to predict the presence of histopathological abnormalities in organs of human or animal subjects based on clinical pathology data, respectively.

[0046] FIG. 2 illustrates a computer-implemented method for generating a machine learning model that can be used to predict the presence of histopathological abnormalities in organs of human or animal subjects based on clinical pathology data. In a first step, training data 2066 is received by analysis system 200. The training data 2066 may be received from a suitable source, for example, from one or more external databases (not shown, but non-limiting examples are provided in the Experimental Results section). Then, in step S101, preprocessing of the training data is performed by preprocessing module 2041 of processor 204. This may include data normalization, as outlined elsewhere in this application, to reduce both inter- and intra-dataset variability. The overall goal of the training scheme illustrated in FIG. 2 is to train multiple machine learning models using multiple different methods, evaluate the performance of each trained machine learning model, and select the best-performing trained machine learning model. Thus, in step S102, a (first) machine learning model is trained using a first method using training module 2042. This can mean several things: for example, the machine learning model may be trained using a first type of training algorithm (e.g., gradient boosting algorithm 2062 or random forest algorithm 2064), or the machine learning model may be trained using a first subset of data (e.g., subset 20662 or subset 20664). After the first machine learning model is trained, in step S104, it is determined whether all machine learning models have been trained by the training module 2042. Here, "all" of the machine learning models refers to training the machine learning models in each of the different ways in which they should be trained. This may include, for example, training the machine learning models using all combinations of subsets 20662, 20664 of the training data 2066 and training algorithms (e.g., gradient boosting algorithm 2062 and random forest algorithm 2064).If it is determined in step S104 that all machine learning models have not been trained, the computer-implemented method returns to step S102, where an additional machine learning model is trained in a different manner than the first. This iterative procedure continues until it is determined in step S104 that all machine learning models have been trained. The computer-implemented method then proceeds to step S106, where a testing module 2044 evaluates the performance of each of the trained machine learning models. Preferably, the evaluation by the testing module 2044 involves determining or calculating a performance metric, as outlined elsewhere in this application. The testing module 2044 may be configured to apply each trained machine learning model to test data 2068 and determine or calculate a performance metric based on the results of applying the trained machine learning model to the test data 2068. The test data 2068 may be a subset of the training data 2062 that was not used to train the machine learning models (to avoid training bias). In the example of FIG. 2, the performance evaluation occurs after all machine learning models have been trained. However, it is equally feasible to perform performance evaluation of a given machine learning model immediately after the machine learning model is generated. Alternatively, the performance evaluation process and the training process may occur in parallel. Other configurations are also contemplated. After the performance of all of the trained machine learning models is evaluated by the testing module 2044 in step S106, the computer-implemented method proceeds to step S108, where the selection module 2046 selects the trained machine learning model with the best performance. This is preferably done according to the value of the performance metric. The selected machine learning model 2072 may then be stored in the memory 206 of the analysis system 200.

[0047] FIG. 2 relates to the generation of a machine learning model. FIG. 3 shows a computer-implemented method according to a first aspect of the present invention, in which the presence or absence of a histopathological abnormality is predicted. In a first step S200, clinical pathology data 20702 is received. For example, the processor 204 of the analysis system 200 may receive the clinical pathology data 20702 from the data acquisition unit 100. For processing, the clinical pathology data 20702 may be stored in a buffer 2070 of the memory 206 of the analysis system 200. Subsequently, in step S202, the analysis model 2048 may retrieve the trained machine learning model 2072 from the memory 206 of the analysis system and apply it to the clinical pathology data 20702. As explained, the trained machine learning model 2072 is preferably configured to output a binary histopathological score indicating the presence or absence of a histopathological abnormality based on the clinical pathology data 20702. The result is then output in step S204. This may include the processor 204 sending the results to the display module 300, where they may be displayed to, for example, a clinician or other client.

[0048] Experimental results Here, we describe the results of a study demonstrating the effectiveness of the computer-implemented method of the present invention. The experiment focused on the detection of histopathological lesions in the livers of rats in toxicity studies. The data included in the experimental study were from rats for which both clinical pathology data and histopathological evaluation of the liver were available. Data from intermediate bleeding points not related to histopathology, as well as data from recovery animals, were excluded.

[0049] The analysis was performed using three datasets (i.e., three sets of training data): -(i) TG-GATE: An online open-source dataset of toxicogenomics studies in rats, including whole slide images (20x) and associated clinical pathology data and histopathology findings, covering classic test items with well-characterized hepatotoxicity. The final dataset for liver cases consisted of 3,894 individual animal entries. Only male rats were present in this data collection. - (ii) Roche Toxicogenomics Initiative: The rat dataset represents the Toxicogenomics Initiative conducted between 1999 and 2003. The test items used in these studies were classical hepatotoxic or nephrotoxic drugs. Hepatotoxic drugs induced toxicity by the following mechanisms: direct action, steatosis, cholestatic, or immune-mediated. Several Roche compounds that were withdrawn due to clinical hepatotoxicity were also included. The dataset consisted of 1,768 individual entries, 1,705 males and 63 females. - (iii) Roche Toxicology Studies: 600 completed or SEND-like (Standardization for Exchange of Nonclinical Data) studies over the past 20 years. The dataset spanned 20 years and consisted of 43,878 entries extracted from over 600 studies. Of this larger dataset, a smaller section (called (iv)) used in another project was extracted and tested separately. This extracted dataset consisted of 832 entries extracted from 34 studies.

[0050] As described below, different combinations of the above datasets were also tested to assess their impact on the AUC of machine learning models generated using the different datasets.

[0051] All data sets were visually inspected using the data analysis software Spotfire (TIBCO).

[0052] Earlier in this application, we mentioned the use of data normalization to reduce variability. In these experiments, a variety of different normalization techniques were tested to minimize intra- and inter-dataset (batch effect) variability, including log-normalization, batch normalization, z-score normalization, and location scale models.

[0053] Here, we consider the location scale model reference range. Due to the diversity of data sets in terms of both time frame and origin, normalized reference values were collected from the literature. Among the various different location scale models, we chose the Chuang-Stein model (https: / / www.lexjansen.com / phuse / 2019 / dh / DH05.pdf). The concept of the Chuang-Stein model is to normalize all values with respect to a selected set of reference ranges.

number

[0054] where: -value is the number to be normalized. -ULN and LLN are the upper and lower limits, respectively, within the dataset for the variable of interest. -ULN std and LLN std are the upper and lower limits, respectively, within the standard criteria list for the variable of interest.

[0055] Reference values were collected from the literature. Male reference values were considered in this particular context due to the higher prevalence of male subjects compared to female subjects in the dataset. To obtain comparable data, the same reference interval was applied to all datasets tested.

[0056] For all datasets, missing values were replaced with randomly generated values.

[0057] Depending on the type of dataset, the histopathological score was calculated in two ways. For the Roche Toxicogenomics Initiative study, a senior toxicological pathologist manually reviewed histological diagnosis and severity to obtain a final score. As a general approach, controls were set as normal (0) unless specific relevant findings were detected (e.g., liver necrosis). Treated animals were evaluated based on their histological score; samples with a score of 1 or less out of 5 were set as normal (0). Samples with a histological score of 2 or more out of 5 were set as pathological (1). The exception to this rule was liver glycogenosis, which was always considered normal. For the Roche toxicology study and TG-GATE, a Python script analyzed the severity of findings reported by pathologists and assigned an automatic score. The script was configured to consider findings marked as "not remarkable," "present," or with a severity of 1 out of 5 as normal (represented as 0 in the table). Results with higher severity were set as pathological (1). A junior toxicological pathologist manually reviewed the automatic scores to make them similar to the scores given by senior toxicological pathologists in toxicogenomics.

[0058] The data was then processed using: 1. Random forest and gradient boosting implementation in Orange (https: / / orangedatamining.com / ). 2. Random Forest implementation in Python using scikit-learn (https: / / scikit-learn.org / stable / modules / generated / sklearn.ensemble.RandomForestClassifier.html).

[0059] Each dataset was partitioned by study ID to avoid exposing the target algorithm to prior study information. Histopathological scores were used as the target value.

[0060] The settings for Orange are shown below. [Table 1]

[0061] The settings in the Python implementation were as follows: [Table 2]

[0062] High variability was observed within and between datasets. Notably, high variability was observed between TG-GATE and toxicogenomics, but lower variability was observed between TG-GATE and Roche general studies and the Roche Toxicogenomics Initiative and Roche toxicology studies. Within-dataset variability was particularly high among the toxicogenomics studies.

[0063] After normalization, the data distribution was similar among the three data sets, as shown in Tables 1, 2 and 3 in the appendix to this patent application.

[0064] Among all types of data normalization, the location-scale model has been an effective technique for normalizing clinicopathological data.

[0065] The results for each dataset processed alone with Orange are shown below, with the datasets having the same labels assigned above. [Table 3] 1.CA=classification accuracy 2. F1 = harmonic mean of precision and recall. 3.https: / / en.wikipedia.org / wiki / Precision_and_recall

[0066] The Roche Toxicogenomics Initiative (ii) dataset was combined with other datasets and processed with Orange, with the following results: [Table 4]

[0067] When data set (iv) was combined with TG-GATE (i), the results for Orange were as follows: [Table 5]

[0068] The Roche toxicity dataset (iii) was processed with Orange in combination with TG-GATE (i). [Table 6]

[0069] Dataset (iv) was not included in the study as it is a subset of the Roche toxicity dataset. Results for 10-fold cross-validated random forest: [Table 7]

[0070] Regardless of the number of samples, a high degree of predictive variability was observed between data sets. According to our preliminary results, Roche toxicity data performed better when combined with TG-GATE. The combination of Roche toxicity and TG-GATE data sets maintained an AUC of 0.82, demonstrating that clinical pathology can be used to detect liver histopathological lesions.

[0071] General Statements The features disclosed in the foregoing description, or in the following claims, or in the accompanying drawings, and expressed in a specific form or in terms of means for performing a disclosed function or methods or processes for obtaining a disclosed result, may be utilized, individually or in any combination of such features, as appropriate, to realize the invention in diverse forms thereof.

[0072] While the present invention has been described in conjunction with the foregoing exemplary embodiments, many equivalent modifications and variations will be apparent to those skilled in the art given this disclosure. Accordingly, the exemplary embodiments of the invention described above are considered to be illustrative and not limiting. Various changes may be made to the described embodiments without departing from the spirit and scope of the invention.

[0073] For the avoidance of doubt, any theoretical explanations provided herein are provided for the purpose of enhancing the understanding of the reader, and the inventors do not wish to be bound by any of these theoretical explanations.

[0074] Any section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.

[0075] Throughout this specification, including the claims which follow, unless the context requires otherwise, the words "comprise" and "include", as well as variations such as "comprises", "comprising", and "including", are to be understood as meaning the inclusion of a stated integer or step or group of steps, but not the exclusion of any other integer or step or group of steps.

[0076] It must be noted that, as used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from "about" one particular value and / or to "about" another particular value. When such a range is expressed, another embodiment includes from the one particular value and / or to the other particular value. Similarly, when values are expressed as approximations, by use of the antecedent "about," it will be understood that the particular value forms another embodiment. The term "about" with respect to numerical values is arbitrary and means, for example, + / - 10%. [Table 8] [Table 9] [Table 10]

Claims

1. A computer implementation method for predicting the presence of histopathological abnormalities in the organs of a human or animal subject based on clinical pathological data, Receiving clinical and pathological data obtained from human or animal subjects, Applying an analytical model to the aforementioned clinical pathology data, wherein the analytical model is configured to output results indicating the possibility of the presence of histopathological abnormalities in the organs of the human or animal subject, Outputting the results, Computer implementation methods including

2. The computer implementation method according to claim 1, wherein the clinical and pathological data includes the ratio of the concentration of albumin in the body fluid of the human or animal subject to the concentration of globulin in the body fluid of the human or animal subject.

3. The computer implementation method according to claim 1, wherein the clinical and pathological data includes one or more concentration measurements of one or more liver injury biomarkers in the body fluids of the human or animal subject.

4. The computer implementation method according to claim 3, wherein the one or more liver injury biomarkers include bilirubin, aspartate aminotransferase, gamma glutamine transferase, alanine aminotransferase, and lactate dehydrogenase.

5. The computer implementation method according to claim 1, wherein the clinical and pathological data includes measured values ​​of the concentration of creatinine kinase in the body fluids of the human or animal subject.

6. The computer implementation method according to claim 1, wherein the clinical and pathological data includes measured values ​​of potassium concentration in the body fluids of the human or animal subject.

7. The computer implementation method according to claim 1, wherein the clinical and pathological data does not include image data.

8. The computer implementation method according to claim 1, wherein the analysis model is a machine learning model trained to output results indicating the possibility of the presence of histopathological abnormalities in the organs of the human or animal subject, based on inputs including clinical and pathological data obtained from the human or animal subject.

9. The computer implementation method according to claim 8, wherein the machine learning model is a random forest algorithm or a gradient boosting algorithm.

10. The computer implementation method according to claim 1, wherein the analysis model is configured to output a histopathological score indicating the possibility of the presence of histopathological abnormalities in the organs of the human or animal subject.

11. The computer implementation method according to claim 10, wherein the histopathological score is a binary score.

12. The computer implementation method according to claim 1, wherein the organ of the human or animal subject is the liver.

13. The computer implementation method according to claim 1, wherein the histopathological abnormality is a lesion.

14. A computer implementation method for generating a machine learning model configured to output results indicating the possibility of the presence of histopathological abnormalities in the organs of a human or animal subject, Receiving training data for each of multiple human or animal subjects, including clinicopathological data and histopathological scores indicating whether histopathological abnormalities are present in the organs of that human or animal subject, Training the machine learning model using the received training data, Computer implementation methods including

15. The computer implementation method according to any one of claims 1 to 13, wherein the analysis model is a machine learning model trained according to the computer implementation method described in claim 14.

Images