Detecting anomalies in specimen images

Abstract

A computer-implemented method for detecting the presence of morphologically abnormal cells in a specimen image includes receiving electronic image data representing a specimen image, the specimen image depicting a plurality of cells; applying an analytical model to the image data of each subset of the plurality of subsets, each subset corresponding to a respective portion of the specimen image depicting a single cell, the analytical model being configured to output, for each subset of image data, a value parameterizing a characteristic of the cell and either a confidence score or an uncertainty score associated with the value, thereby generating output data including a plurality of confidence scores or a plurality of uncertainty scores; and determining, based on the output data, whether one or more morphologically abnormal cells are likely to be present in the specimen image.

Description

[Technical Field]

[0001] The present invention relates to a computer-implemented method and associated system for detecting the presence of morphologically abnormal cells in images of specimens. [Background technology]

[0002] To train a deep neural network to recognize entities within images and accurately identify them, training data (consisting of cell images and type labels) is required to establish ground truth.

[0003] In the case of hematology, there is a particular interest in distinguishing between healthy white blood cells (normal cells) and unhealthy white blood cells (abnormal cells, which may be malignant or benign).

[0004] While it is possible to collect sufficient data from common abnormalities (e.g., common leukemia, infectious diseases), it is not possible to collect sufficient data from rare diseases. Rare conditions can result in significantly different cell appearances, and as a result, neural networks that have not been trained on these rare appearances will not be able to classify cells with high accuracy.

[0005] The present invention addresses these problems. Summary of the Invention

[0006] At a high level, the present invention provides a computer-implemented method for detecting the presence of a population of cells with morphological abnormalities by detecting the presence of a threshold number of cells in a specimen image that cannot be identified with a predetermined confidence score or that cannot be identified below a predetermined uncertainty score. An increase in the population of cells in a sample that cannot be easily identified is an indication that cells with morphological abnormalities are present. The computer-implemented method provided by the present invention is advantageous because it is intended to be able to detect the presence of a population of morphologically abnormal cells without the need to train an analytical model (e.g., a machine learning model) on data outside the domain. The computer-implemented method of the present invention also does not require any modifications to existing "vanilla" models.

[0007] More specifically, a first aspect of the present invention provides a computer-implemented method for detecting the presence of morphologically abnormal cells in a specimen image, the computer-implemented method comprising: receiving electronic image data representing a specimen image, the specimen image depicting a plurality of cells; applying an analytical model to the image data of each subset of a plurality of subsets, each subset corresponding to a respective portion of the specimen image depicting a single cell, the analytical model being configured to output, for each subset of image data, a value parameterizing a characteristic of the cell and either a confidence score or an uncertainty score associated with the value, thereby generating output data including a plurality of confidence scores or a plurality of uncertainty scores; and determining, based on the output data, whether one or more morphologically abnormal cells are likely to be present in the specimen image.

[0008] As used herein, "may be present" may be understood to mean "present." In some cases, rather than making an active determination regarding the likelihood of morphologically abnormal cells being present, the computer-implemented method may include determining whether predetermined abnormality criteria are met. If the predetermined abnormality criteria are met, this may indicate that morphologically abnormal cells are likely to be present or that morphologically abnormal cells are present. After determining whether morphologically abnormal cells are present or whether the predetermined abnormality criteria are met, the computer-implemented method may further include generating an output indicating the result of the determination. Generating the output may include generating instructions that, when executed by a display component of the computing device, cause the display component to display a visual indication of the result of the determination. Alternatively, or additionally, the computer-implemented method may include transmitting the output to a database accessible by a clinical computer system.

[0009] If the criteria for abnormality are determined to be met or if the presence of morphologically abnormal cells is determined to be possible, the output preferably includes a flag. The flag may be added to the electronic image data representing the specimen image. The flag does not necessarily specifically indicate the presence of morphologically abnormal cells in the specimen image, but rather flags to the clinician that the specimen image requires further attention.

[0010] However, in a clinical setting, it is desirable for a variety of reasons to simply flag the results to the clinician, rather than making an automated diagnosis. The computer-implemented methods of the invention may include generating a flag in response to a determination that one or more morphologically abnormal cells may be present.

[0011] In the context of this application, the term "value" may refer not only to a numerical value, but also to a non-numeric value, such as a numerical form indicating a particular classification, or a classification output (more on this further below), which may alternatively be in the form of a text output.

[0012] As used herein, a "morphologically abnormal cell" is a cell whose physical structure differs from that of a normal cell, such that its appearance in an image of a specimen differs from that of a normal or "healthy" cell. Many diseases or other conditions can be detected by the presence of such morphological abnormalities. Thus, the present invention can be used to detect or aid in the detection of those conditions that cause morphological abnormalities. However, it should be understood that the present invention is unlikely to be applied to the detection of conditions that cause abnormalities other than cellular morphological abnormalities. The detection of such conditions is beyond the scope of this patent application.

[0013] In the context of the present invention, a "specimen image" is an image depicting human or animal tissue. The image may be obtained, for example, from a microscope or other imaging device. The specimen image is described as depicting multiple cells. Therefore, the specimen image is preferably at a magnification that allows cells to be individually resolved, for example, by an image processing algorithm. The specimen image may be an image of a slide of a sample of bodily fluid or tissue obtained from a human or animal subject. The bodily fluid may be blood, and therefore the abnormality may be a hematological abnormality. Blood is not the only bodily fluid for which the computer-implemented method of the present invention can be used to achieve clinically meaningful results. For example, the bodily fluid may include a sample of the subject's tissue / organ and / or a sample of a product produced by the subject's tissue / organ. The product produced by the subject's tissue / organ may be, for example, a secretion (e.g., glandular secretion, milk, colostrum, tears, saliva, sweat, pigment, mucus), sputum, semen, vaginal / cervical fluid, blood (plasma, serum), cerebrospinal fluid (CSF), excreta, feces, or a product of urine, skin, or hair.

[0014] The computer-implemented method of the first aspect of the present invention includes applying an analytical model. As used herein, the term "analytical model" may refer to a mathematical model configured to determine at least one target variable for at least one condition variable. The term "target variable" may refer to a clinical value to be predicted. The value of the target variable to be predicted may depend on the disease whose presence or condition is to be predicted. The target variable may be numeric or categorical. For example, the target variable may be categorical, such as "positive" for the presence of disease or "negative" for the absence of disease. In other cases, the target variable may refer to a classification of a cell type (more on this below). The target variable may be at least one value and / or numeric, such as a scale value.

[0015] 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. In other cases, the analytical model may be a segmentation model, the output of which is, for example, a representation of the segment of the sample image to which each pixel is associated.

[0016] In the context of this application, the term "classification model" may be used to refer to an analytical model whose output is a respective classification or score indicating the type of cell depicted in each subset of image data. For completeness, it is noted that when the analytical model is a classification model, the "values ​​parameterizing the characteristics of the cells" may be either a numeric or textual output indicating the type of cell, i.e., the "characteristic" of the cell is the cell type.

[0017] Specifically, the analytical model may be a machine learning model trained to output a respective result indicating the type of cells depicted in the image data of a given subset. The machine learning model may be a regression model or a classification model, as previously defined. The machine learning model is preferably trained using supervised learning. The machine learning model may include a neural network, such as a convolutional neural network.

[0018] In preferred embodiments, the analytical model is a classification model based on a neural network (such as a convolutional neural network and / or a deep neural network). In these cases, the classification model is preferably trained to classify cells as one of multiple cell types based on data representing electronic images of the cells. For example, the classification model may be configured to classify cells as one of multiple types of white blood cells, such as at least one of the following: neutrophils, lymphocytes, monocytes, eosinophils, and basophils. The classification model is preferably trained using training data including multiple images of normal or healthy cells, each image associated with a label indicating the cell type. This demonstrates that the classification model (or more generally, the analytical model) does not require training on morphologically abnormal cells.

[0019] The analytical model may include multiple analytical sub-models, each of which may take the form of a machine learning model as outlined in the previous two paragraphs.

[0020] The concepts of confidence score and uncertainty score are central to the present invention. This term is well known in the fields of statistics and, indeed, machine learning. The present invention is applicable in all cases where an analytical model can establish the distribution of normal samples (i.e., healthy or normal cells). This is because knowing the distribution makes it possible to calculate an uncertainty score or confidence score. In fact, many analytical models are able to calculate an uncertainty or confidence score in addition to the usual output value.

[0021] Different analytical models, such as machine learning models, report different uncertainties. In the computer-implemented method according to the present invention, we are interested in prediction uncertainty caused by the fact that a data point is too far from the training data to be reliably classified, known as epistemic uncertainty (What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? By Kendall & Gal 17).

[0022] In general, uncertainty can be quantified as the amount of discrepancy within the predictive distribution reported by the model. Variance or standard deviation are common ad-hoc choices for this. A theoretically well-founded approach is to define uncertainty as the reciprocal information between model parameters and outputs (Deep Deterministic Uncertainty: A Simple Baseline by Mukhot et al. 2022).

[0023] The predictive distribution can be approximated by using an ensemble of models (“Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift” by Ovadia et al. 2019). Using an ensemble of models can generally refer to the process of combining models trained to perform the same task (in this case, classifying images with the same class). Generally, using an ensemble to combine the outputs of different models is typically done to increase accuracy, as uncorrelated errors introduced by different models cancel out when the outputs are properly combined. However, in this case, the ensemble can be used to obtain a measure of disagreement between the various submodels of the ensemble. See below. If the ensemble predictions of the individual models in the ensemble disagree greatly, this indicates that the data point is far from the training data and therefore should be assigned a high epistemic uncertainty. Conversely, if the models in the ensemble agree, this indicates that the data point is close to the training set and therefore has a low epistemic uncertainty.

[0024] More specifically, in this context, the use of an ensemble may refer to a process in which several analytical submodels are each used to obtain a respective output. The outputs may then be combined, and the variance or standard deviation (or other information-theoretic measure) of the outputs may be used as an uncertainty score. More specifically, applying an analytical model to the image data of each subset of the plurality of subsets may include applying a plurality (or ensemble) of analytical submodels to each subset of image data. Each analytical submodel of the plurality of analytical submodels may then be configured (for each subset's image data) to output a respective plurality of values ​​that each parameterize a characteristic of the cell. Applying an analytical model to the image data of each subset of the plurality of subsets may further include, for each of the outputted plurality of values, determining a statistical characteristic that indicates the reliability or uncertainty of the plurality of values. The statistical characteristic may be the variance or standard deviation, or another information-theoretic parameter. Each of the analytical submodels may be of the same type and may be trained with the same training data. To ensure that the ensemble of analytical sub-models is not all identical (in which case it would be impossible to obtain a meaningful uncertainty score), each analytical sub-model of the multiple analytical sub-models preferably differs at least by an initial random seed that is randomly used to initialize the model before training.

[0025] In some cases, the various analytical sub-models may not be of the same type; for example, some of the analytical sub-models may be convolutional neural networks, some may be transformers, etc. The different types of sub-models may be trained with the same or different training datasets. However, the different types of sub-models are all trained on the same task (i.e., classification into the same classes).

[0026] The predictive distribution can also be approximated by other means, for example, by using dropout. In the context of machine learning, the term "dropout" is used to refer to a process in which some number of neurons are randomly ignored (i.e., dropped out). Dropout as well as ensembles can be used as approximations for Bayesian neural networks (Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning by Yarin Gal, Zoubin Ghahramani 2015).

[0027] If the uncertainty per data point is not precise enough, the model can determine whether it is reliable enough to make a decision based on a semantically meaningful collection of data points (such as images of cells from the same sample). In the case of hematology, if the model reports elevated uncertainty for enough cells on a slide, a decision is made as to whether the sample should be flagged.

[0028] We next consider how uncertainty / confidence scores can be used to detect the presence of morphologically abnormal cells. A core concept of the present invention is the ability to use uncertainty or confidence scores to identify a population of morphologically abnormal cells, even in the absence of training data covering the morphologically abnormal cells. The output data includes either multiple confidence scores or multiple uncertainty scores, which can be calculated using the process described in the previous paragraphs. It has been observed that in specimen images depicting samples containing morphologically abnormal cells, there is an increased number or percentage of decreased confidence scores or increased uncertainty scores. This is because analytical models have not been trained to generate output values ​​based on such morphologically abnormal cells.

[0029] In some cases, the computer-implemented method may include detecting that one or more subsets of image data represent morphologically abnormal cells based on a respective uncertainty score or confidence score calculated for each of the one or more subsets of image data. For example, the computer-implemented method may include determining, for a given subset of data, whether the uncertainty score generated for the image data of the subset exceeds a predetermined maximum uncertainty threshold. If the uncertainty score is determined to exceed the predetermined maximum uncertainty threshold, the computer-implemented method determines that the cells represented by the image data of the given subset are morphologically abnormal. Similarly, the computer-implemented method may include determining, for a given subset of data, whether the confidence score generated for the image data of the subset is less than a predetermined minimum confidence threshold. If the uncertainty score is determined to be less than the predetermined minimum confidence threshold, the computer-implemented method determines that the cells represented by the image data of the given subset are morphologically abnormal.

[0030] If a human or animal subject has a condition that leads to the presence of morphologically abnormal cells, the specimen image may contain multiple such cells. Thus, in addition to detecting morphological abnormalities based on confidence or uncertainty scores associated with individual cells, statistical approaches may also be used in which a histogram of uncertainty or confidence scores is considered for the entire population of cells depicted in the specimen image. In these examples, the computer-implemented method may detect the presence of a subset of morphologically abnormal cells without necessarily identifying the specific cells that exhibit the abnormality.

[0031] Therefore, determining whether an abnormality is likely to exist may include determining a proportion of confidence scores in the output data that are less than a predetermined minimum confidence threshold, and determining whether one or more morphologically abnormal cells are likely to exist based on the proportion of confidence scores that are less than the predetermined minimum confidence threshold. Thus, the computer-implemented method may include determining a proportion of confidence scores in the output data that are less than a predetermined minimum confidence threshold, and determining whether one or more morphologically abnormal cells are likely to exist based on the proportion of confidence scores that are less than the predetermined minimum confidence threshold. Determining the proportion of confidence scores in the output data that are less than a predetermined confidence threshold may include, for each subset of data, determining whether the confidence score is less than or equal to a predetermined minimum confidence threshold, counting the number of data in the subset that have confidence scores that are less than or equal to the predetermined minimum confidence threshold, and dividing the counted number by the total number of data in the subset. Determining whether one or more morphologically abnormal cells are likely to be present based on the percentage of confidence scores that are less than a predetermined minimum confidence threshold may include determining whether the percentage of confidence scores exceeds a predetermined threshold percentage, and determining that there is a possibility of the presence of morphologically abnormal cells in the specimen image if the percentage of confidence scores is determined to exceed the predetermined threshold percentage.

[0032] A similar process can be performed using uncertainty scores rather than confidence scores.

[0033] Specifically, determining whether an abnormality is likely to be present may include determining a proportion of uncertainty scores in the output data that exceed a predetermined maximum uncertainty threshold, and determining whether one or more morphologically abnormal cells are likely to be present based on the proportion of uncertainty scores that exceed the predetermined maximum uncertainty threshold. Thus, the computer-implemented method may include determining a proportion of uncertainty scores in the output data that exceed a predetermined maximum uncertainty threshold, and determining whether one or more morphologically abnormal cells are likely to be present based on the proportion of uncertainty scores that exceed the predetermined maximum uncertainty threshold. Determining a proportion of uncertainty scores in the output data that exceed a predetermined maximum uncertainty threshold may include determining, for each subset of data, whether the confidence score exceeds a predetermined maximum uncertainty threshold, counting the number of subset data whose uncertainty scores exceed the predetermined maximum uncertainty threshold, and dividing the counted number by the total number of subset data. Determining whether one or more morphologically abnormal cells are likely to be present based on the proportion of uncertainty scores that exceed a predetermined minimum confidence threshold may include determining whether the proportion of uncertainty scores exceeds a predetermined threshold percentage, and determining that there is a possibility of morphologically abnormal cells in the specimen image if the proportion of uncertainty scores exceeds the predetermined threshold percentage.

[0034] In some cases, the output may include a display of the subset of data whose uncertainty scores exceed a predetermined maximum uncertainty threshold or whose confidence scores are below a predetermined minimum confidence threshold. The computer-implemented method may then further include instructions that, when executed by a display component of the computing device, display an annotated version of the specimen image. Preferably, the annotated version of the specimen image includes a display (e.g., in the form of an annotation or overlay) of cells whose uncertainty scores exceed a predetermined maximum uncertainty threshold or whose confidence scores are below a predetermined minimum confidence threshold. In this way, the clinician is presented with an ergonomically improved display that highlights cells requiring further attention, ultimately reducing the time required to analyze the specimen image.

[0035] A second aspect of the present invention may provide a clinical support system comprising a processor, the processor configured to execute the computer-implemented method of the first aspect of the present invention. The clinical support system may further include a display component. It will be understood that any feature described above with respect to the first aspect of the present invention applies equally well to the second aspect of the present invention, unless clearly incompatible or the context clearly dictates otherwise. The clinical support system may include appropriate modules configured to perform each of the different operations of the computer-implemented method of the first aspect of the present invention.

[0036] A third aspect of the present invention may provide a computer program comprising instructions that, when executed by a processor of a computer, cause the processor to perform the computer-implemented method of the first aspect of the present invention. It will be understood that any feature described above with respect to the first aspect of the present invention applies equally well to the third aspect of the present invention, unless clearly incompatible or the context clearly dictates otherwise. A fourth aspect of the present invention may provide a computer-readable storage medium having stored thereon a computer program of the third aspect of the present invention. It will be understood that any feature described above with respect to the first aspect of the present invention applies equally well to the fourth aspect of the present invention, unless clearly incompatible or the context clearly dictates otherwise.

[0037] A fifth aspect of the present invention may provide a computer-implemented method for detecting the presence of morphologically abnormal cells in an image of a specimen, the computer-implemented method including: receiving electronic image data representing an image of the specimen, the image of the specimen depicting a plurality of cells; applying an analytical model to the image data of each subset of the plurality of subsets, each subset corresponding to a respective portion of the image of the specimen depicting a single cell, the analytical model being configured to output, for each subset of image data, a value parameterizing a characteristic of the cell and either a confidence score or an uncertainty score associated with the value, thereby generating output data including the plurality of confidence scores or the plurality of uncertainty scores; determining a proportion of the confidence scores in the output data that are below a predetermined minimum confidence threshold or a proportion of the uncertainty scores in the output data that exceed a predetermined maximum uncertainty threshold; determining whether the proportion exceeds a predetermined threshold proportion; and generating an output indicative of the result of the determination, the output including a flag if the proportion is determined to exceed the predetermined threshold proportion. The flag preferably alerts a clinician to the possible presence of morphologically abnormal cells. It will be understood that any feature described above in relation to the first aspect of the invention applies equally well to the fifth aspect of the invention, except where clearly incompatible or where the context clearly dictates otherwise.

[0038] A sixth aspect of the present invention may provide a computer program comprising instructions which, when executed by a processor of a computer, cause the processor to perform the computer-implemented method of the fifth aspect of the present invention. It will be understood that any feature described above with respect to the first aspect of the present invention applies equally well to the sixth aspect of the present invention, unless clearly incompatible or the context clearly dictates otherwise. A seventh aspect of the present invention may provide a computer-readable storage medium having stored thereon a computer program of the sixth aspect of the present invention. It will be understood that any feature described above with respect to the first aspect of the present invention applies equally well to the seventh aspect of the present invention, unless clearly incompatible or the context clearly dictates otherwise.

[0039] An eighth aspect of the present invention may provide a clinical support system comprising a processor, the processor configured to execute the computer-implemented method of the fifth aspect of the present invention. The clinical support system may further include a display component. Any features described above with respect to the first and fifth aspects of the present invention apply equally well to the eighth aspect of the present invention, unless clearly incompatible or the context clearly indicates otherwise. The clinical support system may include appropriate modules configured to perform each of the different operations of the computer-implemented method of the fifth aspect of the present invention.

[0040] 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]

[0041] Embodiments of the present invention will now be described with reference to the accompanying drawings.

[0042] [Figure 1] 1 illustrates a clinical support system configured to perform a computer-implemented method according to the present invention. [Figure 2]1 is a flowchart illustrating an example of a computer-implemented method according to the present invention. [Figure 3] 1 is a histogram showing the distribution of uncertainty in a population of normal cells. [Figure 4A] 1 is a histogram showing the distribution of uncertainty in a population of cells containing immature granulocytes. [Figure 4B] 1 is a histogram showing the distribution of uncertainty in a population of cells containing abnormal blast cells. DETAILED DESCRIPTION OF THE INVENTION

[0043] Aspects and embodiments of the present invention will now be described with reference to the accompanying figures. Further aspects and embodiments will be apparent to those skilled in the art. All documents mentioned herein are incorporated by reference.

[0044] FIG. 1 is a schematic diagram of a clinical support system 1 according to, for example, a first embodiment of the present invention. The clinical support system 1 includes a processor 12, a memory 14, and a display component 16. While these components are shown in FIG. 1 as being all part of the same system, it will be understood that the system may be a distributed system in which various components are located on different hardware, optionally in different locations. In such cases, the components (e.g., the processor 12, the memory 14, and the display component 16) may be connected via a network (not shown). The network 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. The structure of the clinical support system 1 will now be described with reference to FIGS. 2 and 4 before discussing the operations it is configured to perform. The processor 12 includes multiple modules. In this specification, the term "module" is used to refer to a functional module configured or adapted to perform a specific function. The modules may be implemented in hardware (i.e., they may be separate physical components within a computer), software (i.e., they may represent separate sections of code that, when executed by the processor 12, cause it to perform particular functions), or a combination of both.

[0045] Specifically, processor 12 of Figure 1 includes a cell identification module 120, an analysis module 122, an uncertainty determination module 124, an anomaly detection module 126, and an output module 128. The functionality of each of these modules will be described in more detail shortly. Memory 14 may be in the form of permanent memory or temporary memory, or may include a combination of the two. Memory 14 stores an analytical model 140 and a set of thresholds 142. Display component 16 is preferably in the form of a VPU, screen, or monitor configured to visually render the data to a clinician for viewing results.

[0046] 2 illustrates a process performed by the processor 12 of the clinical support system 1. In a first step S200, image data is received by the processor 12. The image data, more specifically, is electronic image data representing an image of a specimen, the specimen image depicting a plurality of cells. The electronic image data will naturally include a plurality of subsets of image data, each subset corresponding to a respective portion of the image data depicting a single cell. In step S202, the cell identification module 120 is configured to identify the subset of data corresponding to each individual cell. This may be done by applying an image analysis algorithm to the electronic image data received in step S200. The output of step S202 is a plurality of subsets of image data, each subset corresponding to a portion of the specimen image depicting a single cell.

[0047] Steps S204 and S206 are performed on all subset image data output by step S202. However, for brevity, only the process for a single subset of data will be described. In step S204, an analytical model 140 is retrieved from memory 14 and applied to the subset image data by analysis module 122. The output of step S204, for each subset of data, is a value that parameterizes a characteristic of the cell represented by the subset image data. For example, analytical model 140 may be in the form of a classification model, and the output of step S204 is a classification of the type of cell depicted. In step S206, uncertainty determination module 124 determines, for each subset of image data, an uncertainty value associated with the output generated by analysis module 122. Various methods can be used to determine uncertainty. Note that in other embodiments, confidence scores can be generated rather than uncertainty scores. The output of step S206 is an uncertainty score for each subset of image data. Collectively, this may be referred to as output data.

[0048] In step S208, the anomaly detection module 126 of the processor 12 determines whether morphologically abnormal cells may be present in the specimen image based on the output data. While there are various ways to do this, in one embodiment, the percentage of cells whose uncertainty score exceeds a predetermined maximum uncertainty threshold is determined. This percentage is then compared to a predetermined threshold percentage, and if it exceeds the predetermined threshold percentage, this indicates the presence of one or more morphologically abnormal cells in the specimen image. This is because high uncertainty is associated with unknown cell types for which the analytical model is not or insufficiently trained due to a lack of training data (so-called "epistemic uncertainty"). Again, it will be understood that a similar procedure can be performed using confidence scores instead of uncertainty scores.

[0049] After the determination is made by the uncertainty determination module 126 in step S208, an output is generated by the output module 128 in step S210. In some cases, the output may be transmitted to a database where it can be accessed by a clinical computing network. Alternatively, the output may be a visual output that may include a flag, as described earlier in this application.

[0050] Experimental results We were able to test the present invention using a classification model configured to identify white blood cells. A training set of eight (normal) specimen slides was used, and a validation set of four specimen slides was used. Approximately 600 images were derived from each slide. For abnormal examples, 10 specimen images containing immature granulocytes and 13 specimen images containing blast cells were used. An illustrative table is provided below. TIFF2025536473000001.tif57121

[0051] Abnormal slides were used for validation purposes only.

[0052] An ensemble of five convolutional neural networks was trained on normal specimen images to classify cells into normal types: neutrophils, lymphocytes, monocytes, eosinophils, basophils, and immature granulocytes. An overall F1 score of 92% was achieved on the validation set (five normal classes + immature granulocytes). Due to the small number of normal slides, the validation set was also used for testing.

[0053] After training, the ensemble predictions of all abnormal and normal (validation set) slides were used to approximate the uncertainty. Histograms of the uncertainty distribution were used to distinguish normal from abnormal samples. The histograms are shown in Figures 3, 4A, and 4B.

[0054] Figure 3 shows normal samples. The x-axis represents the uncertainty value, and the y-axis represents the number of cells in the specimen image at the corresponding uncertainty level. Figure 4A shows a histogram representing the distribution of uncertainty in a sample containing morphologically abnormal immature granulocytes, and Figure 4B shows a histogram representing the distribution of uncertainty in a sample containing morphologically abnormal blast cells. Note that in both cases, the scale of the x-axis is different from that in Figure 3A. It is clear that the number of cells with higher uncertainty scores is significantly higher in Figures 4A and 4B than in Figure 3A, demonstrating that it is possible to detect samples with morphologically abnormal cells from consideration of the uncertainty or confidence score.

[0055] General description 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 a method or process for obtaining a disclosed result, may be utilized, individually or in any combination of such features, as appropriate, to realize the invention in its various forms.

[0056] 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 can be made to the described embodiments without departing from the spirit and scope of the invention.

[0057] 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.

[0058] Any section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described.

[0059] 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 understood to mean the inclusion of a stated integer or step or steps but not the exclusion of any other integer or step or steps.

[0060] 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%.

Claims

1. 1. A computer-implemented method for detecting the presence of morphologically abnormal cells in an image of a specimen, the computer-implemented method comprising: receiving electronic image data representing an image of a specimen, the image of the specimen depicting a plurality of cells; applying an analytical model to the image data of each subset of the plurality of subsets, each subset corresponding to a respective portion of the specimen image depicting a single cell, the analytical model being configured to output, for the image data of each subset, a value parameterizing a property of the cell and either a confidence score or an uncertainty score associated with the value, thereby generating output data including a plurality of the confidence scores or a plurality of the uncertainty scores; determining whether one or more morphologically abnormal cells may be present in the image of the specimen based on the output data; 11. A computer-implemented method comprising:

2. the analytical model is a classification model; The computer-implemented method of claim 1 , wherein the values ​​parameterizing the properties of the cells are numeric or textual outputs indicative of the type of the cells.

3. 3. The computer-implemented method of claim 2, wherein the classification model is a neural network model trained using training data including a plurality of images of normal or healthy cells to classify a cell as one of a plurality of cell types based on data representing an electronic image of the cell, each image being associated with a label indicating the cell type.

4. 4. The computer-implemented method of claim 3, wherein the classification model is configured to classify cells as one of a plurality of types of white blood cells, the plurality of types of white blood cells including neutrophils, lymphocytes, monocytes, eosinophils, and basophils.

5. applying an analytical model to image data of each subset of the plurality of subsets applying a plurality of analytical sub-models to each subset of the image data, each analytical sub-model of the plurality of analytical sub-models configured to output, for each subset of image data, a respective plurality of values ​​each parameterizing a property of the cell; and determining a variance or standard deviation for each of the output values, the determined variance or standard deviation corresponding to the uncertainty score.

4. The computer-implemented method of claim 2 or 3, comprising:

6. 6. The computer-implemented method of claim 5, wherein each of the plurality of analytical sub-models is trained on the same training data using a different initial random seed to initialize the sub-models prior to training.

7. The computer-implemented method of claim 2 , wherein the uncertainty score is calculated using dropout.

8. determining a proportion of the confidence scores in the output data that are below a predetermined minimum confidence threshold; and determining whether one or more morphologically abnormal cells are likely to be present based on the proportion of the confidence scores that are less than the predetermined minimum confidence threshold. The computer-implemented method of claim 1 , further comprising:

9. Determining whether one or more morphologically abnormal cells are likely to be present in the image of the specimen based on the percentage of the confidence scores below the predetermined confidence threshold includes: determining whether a percentage of the confidence scores exceeds a predetermined threshold percentage; and If it is determined that the percentage of the confidence scores exceeds the predetermined threshold percentage, it is determined that there is a possibility of the presence of morphologically abnormal cells in the specimen image. The computer-implemented method of claim 8 , comprising:

10. determining a proportion of the uncertainty scores in the output data that exceed a predetermined maximum uncertainty threshold; and determining whether one or more morphologically abnormal cells are likely to be present based on the percentage of the uncertainty scores that exceed the predetermined maximum uncertainty threshold; The computer-implemented method of claim 1 , further comprising:

11. Determining whether one or more morphologically abnormal cells are likely to be present in the image of the specimen based on the percentage of the uncertainty scores that exceed the predetermined uncertainty threshold includes: determining whether the percentage of the uncertainty scores exceeds a predetermined threshold percentage; and determining that there is a possibility of the presence of morphologically abnormal cells in the specimen image if it is determined that the percentage of the uncertainty scores exceeds the predetermined threshold percentage; The computer-implemented method of claim 10, comprising:

12. 12. The computer-implemented method of claim 1, wherein if it is determined that an anomaly may be present in the specimen image, the computer-implemented method further comprises adding a flag to the electronic image data representing the specimen image.

13. The computer-implemented method of claim 1 , wherein the specimen image is an image of a slide of a sample of a body fluid obtained from a human or animal subject.

14. The computer-implemented method of claim 13 , wherein the bodily fluid is blood.

15. 1. A computer-implemented method for detecting the presence of morphologically abnormal cells in an image of a specimen, the computer-implemented method comprising: receiving electronic image data representing an image of a specimen, the image of the specimen depicting a plurality of cells; applying an analytical model to the image data of each subset of the plurality of subsets, each subset corresponding to a respective portion of the specimen image depicting a single cell, the analytical model being configured to output, for the image data of each subset, a value parameterizing a property of the cell and either a confidence score or an uncertainty score associated with said value, thereby generating output data comprising a plurality of said confidence scores or a plurality of said uncertainty scores; determining the proportion of the confidence scores in the output data that are below a predetermined minimum confidence threshold; or determining a percentage of the uncertainty scores in the output data that exceed a predetermined maximum uncertainty threshold to determine whether the percentage exceeds a predetermined threshold percentage; and generating an output indicative of the result of the determination, wherein if the percentage is determined to exceed the predetermined threshold percentage, the output includes a flag.

11. A computer-implemented method comprising:

Images