Electronic device for pregnancy loss classification and related methods

Abstract

Disclosed is an electronic device for pregnancy loss classification, the electronic device comprising an interface, one or more processors, and a memory. The one or more processors are configured to obtain male data associated with a male; obtain female data associated with a female; obtain pregnancy loss data associated with a fetus and / or fetal tissue; determine one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status; and provide an output including a first output associated with the one or more risk scores.

Description

[0001] ELECTRONIC DEVICE FOR PREGNANCY LOSS CLASSIFICATION AND RELATED METHODS

[0002] The present disclosure pertains to the field of electronic devices, and in particular electronic devices for processing of medical data, systems, and related methods.

[0003] BACKGROUND

[0004] Pregnancy loss is a well-documented occurrence in the modern world. More than one in four pregnancies worldwide end in pregnancy loss. However, the causal factors contributing to pregnancy losses or miscarriages may not be easily identifiable. Often pregnancy losses are not fully investigated due to high costs.

[0005] SUMMARY

[0006] There is a need for tools and methods for improving classification, characterization, identification, and / or prediction of pregnancy loss.

[0007] Accordingly, an electronic device for pregnancy loss classification is provided, the electronic device comprising an interface, one or more processors, and a memory. The one or more processors may be configured to obtain male data associated with a male; optionally obtain female data associated with a female; optionally obtain pregnancy loss data, e.g. associated with a fetus and / or fetal tissue; determine one or more risk scores based on one or more, such as at least two or all, of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and optionally including a first risk score e.g. indicative of a fetal ploidy status; and provide an output including a first output associated with the one or more risk scores.

[0008] Also disclosed is a computer-implemented method for pregnancy loss classification. The method comprises one or more, such as two or all, of obtaining male data associated with a male; obtaining female data associated with a female; and obtaining pregnancy loss data associated with a fetus and / or fetal tissue. The method comprises determining one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and optionally including a first risk score e.g. indicative of a fetal ploidy status; and providing an output including a first output associated with the one or more risk scores. Also disclosed is a computer-implemented method for training a neural network to process as inputs one or more of, such as two or all of, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue, and for providing as output one or more risk scores associated with pregnancy loss is disclosed. The method comprises obtaining, using at least one processor, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue. The method comprises performing, using the at least one processor, a training comprising generating, using the at least one processor and a machine-learning model, risk data based on the male data, the female data, and the pregnancy loss data. The method comprises performing, using the at least one processor, a training comprising obtaining, using the at least one processor, training data. The method comprises performing, using the at least one processor, a training comprising determining, using the at least one processor and one or more loss functions, one or more loss parameters based on the male data, the female data, the pregnancy loss data, and the training data. The method comprises performing, using the at least one processor, a training comprising training, using the at least one processor, the machine learning model based on one or more loss parameters.

[0009] Further, systems, devices, and non-transitory computer readable medium implementing functionality and / or methods as described herein is disclosed.

[0010] The present disclosure allows for more accurate and robust pregnancy loss classification, e.g. by taking into account data associated with the fetus, the female and the male. Fetus, female / female, and male / male are also collectively referred to as the trio.

[0011] Advantageously, the present disclosure may not require invasive procedures, such as an endometrial biopsy, and / or collection of material from the lost fetus. Instead, the present disclosure can allow for pregnancy loss classification to be performed based on data obtained via pregnancy-non-invasive means, such as data obtained via blood and / or urine and / or mikrobiome samples and / or semen.

[0012] Advantageously, the present disclosure incorporates information / data from, indicative of and / or associated with the trio, i.e. fetus, female, and male. Taking the trio (fetus, female, male) into account can increase the performance and accuracy of pregnancy loss classification and / or estimation / prediction of any future pregnancy-related complication. The present disclosure allows early classification and e.g. is applicable from the first pregnancy loss.

[0013] BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The above and other features and advantages of the present disclosure will become readily apparent to those skilled in the art by the following detailed description of examples thereof with reference to the attached drawings, in which:

[0015] Fig. 1 schematically illustrates an example system according to the present disclosure,

[0016] Fig. 2 is a flow chart of an example computer implemented method according to the present disclosure,

[0017] Fig. 3 is a flow chart of an example computer implemented method according to the present disclosure,

[0018] Fig. 4 is a flow chart of an example computer implemented method according to the present disclosure,

[0019] Fig. 5 illustrates an example implementation of a machine learning model according to the disclosure,

[0020] Fig. 6 is an example AUC-RUC curve associated with an example model according to the present disclosure,

[0021] Fig. 7 is an example AUC-PRC curve associated with an example model according to the present disclosure,

[0022] Fig. 8 shows a calibration curve for Cohort 2,

[0023] Fig. 9 shows a calibration curve for Cohort 3,

[0024] Fig. 10 shows feature importance ranked by absolute mean SHAP values for predictive model,

[0025] Fig. 11 shows a Shapley Additive Explanations (SHAP) bar plot,

[0026] Fig. 12 shows a Shapley Additive Explanations (SHAP) bar plot,

[0027] Fig. 13 shows a Shapley Additive Explanations (SHAP) bar plot, and Fig. 14 shows a Shapley Additive Explanations (SHAP) bar plot.

[0028] DETAILED DESCRIPTION

[0029] Various examples and details are described hereinafter, with reference to the figures when relevant. It should be noted that the figures may or may not be drawn to scale and that elements of similar structures or functions are represented by reference numerals throughout the figures. It should also be noted that the figures are only intended to facilitate the description of the examples. They are not intended as an exhaustive description of the disclosure or as a limitation on the scope of the disclosure. In addition, an illustrated example needs not have all the aspects or advantages shown. An aspect or an advantage described in conjunction with a particular example is not necessarily limited to that example and can be practiced in any other examples even if not so illustrated, or if not so explicitly described.

[0030] The figures are schematic and simplified for clarity, and they merely show details which aid understanding the disclosure, while other details have been left out. Throughout, the same reference numerals are used for identical or corresponding parts.

[0031] It is to be understood that a description of a feature in relation to the system / electronic device is also applicable to the corresponding feature in the method(s) of operating a system / electronic device as disclosed herein and vice versa.

[0032] The present disclosure relates to tools and methods for analysis, classification, monitoring, and / or prediction of pregnancy loss and pregnancy complications.

[0033] An electronic device for pregnancy loss classification is disclosed. The electronic device comprises an interface, one or more processors, and a memory. The one or more processors are configured to obtain, e.g. retrieve from a database and / or receive via the interface, clinical data including one or more of, such as two or all of, male data associated with a male; female data associated with a female; and pregnancy loss data, e.g. associated with a fetus and / or fetal tissue and / or pregnancy tissue. The one or more processors are configured to determine one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss. The one or more processors are configured to provide an output including a first output associated with the one or more risk scores. The one or more risk scores may comprise a first risk score, e.g. indicative of a fetal ploidy status. The first output may be associated with, such as representative or indicative of, the first risk score.

[0034] In one or more example electronic devices, the one or more processors are configured to obtain male data associated with a male; obtain female data associated with a female; obtain pregnancy loss data associated with a fetus and / or fetal tissue; determine one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss; and provide an output including a first output associated with the one or more risk scores.

[0035] In one or more examples, to provide an output including a first output associated with the one or more risk scores comprises to display, on a display of the interface, one or more user interface elements indicative of, e.g. showing the value of, one or more risk scores, such as first user interface element indicative of the first risk score. For example, a first primary user interface element may be color-coded based on a first primary risk score, e.g. indicative of possibility of euploid loss and / or a first secondary user interface element may be color-coded based on a first secondary risk score, e.g. indicative of possibility of euploid loss.

[0036] In one or more examples, to provide an output including a first output associated with the one or more risk scores comprises to transmit the one or more risk scores or a subset thereof to a remote server or database, e.g. via a network.

[0037] In one or more examples, to provide an output including a first output associated with the one or more risk scores comprises to store the one or more risk scores or a subset thereof in the memory or a database.

[0038] In one or more example electronic devices, the one or more processors are configured to determine one or more biomarkers based on one or more of the female data, the male data, and the pregnancy loss data. In one or more example electronic devices, to provide an output comprises to provide a biomarker output associated with, such as representative or indicative of, the one or more biomarkers. The biomarker output may comprise a SHAP representation.

[0039] The one or more biomarkers may be indicative of one or more features of the data input to the ML model influencing the first risk score. For example, to determine the one or more biomarkers may be based on a SHAP representation of the input data to the machine learning model. To determine one or more biomarkers may comprise to determine, for each feature or input to the model or for a subset of features, a value, e.g. a SHAP value, indicative of feature importance in the model, optionally followed by a ranking of the values. In other words, to determine one or more biomarkers may comprise to identify the most important features that have led to the first risk score.

[0040] To provide a biomarker output may comprise outputting a SHAP representation of the input data to the ML model. In other words, biomarker(s) may comprise a data identifier indicative of the input data, see e.g. examples in Table 1 , and a value, such as a SHAP value, indicative of feature / data importance in the ML model. The biomarker output may comprise a subset of SHAP values, e.g. the ten or fifteen most important features of data used as input to the model.

[0041] In one or more example electronic devices, the one or more processors are configured to determine one or more female biomarkers based on one or more of the female data, the male data, and the pregnancy loss data, e.g. the female data and / or the male data. In one or more example electronic devices, to provide an output comprises to provide a female output associated with, such as representative or indicative of, the one or more female biomarkers.

[0042] The one or more female biomarkers may be indicative of one or more features of the female data influencing the first risk score. For example, to determine the one or more female biomarkers may be based on a SHAP representation of the female data, e.g. input to the machine learning model. To determine one or more female biomarkers may comprise to determine, for each feature or input to the model or for a subset of features, a value, e.g. a SHAP value, indicative of feature importance in the model, optionally followed by a ranking of the values. In other words, to determine one or more female biomarkers may comprise to identify the most important features that have led to the first risk score.

[0043] To provide a female output may comprise outputting a SHAP representation of the female data input to the ML model. In other words, female biomarker(s) may comprise a female data identifier indicative of the female data, such as maternal age, maternal vitamin E supplement, maternal vitamin D supplement, maternal systolic blood pressure, maternal pulse, and a value, such as a SHAP value, indicative of feature / data importance in the model. The female output may comprise a subset of SHAP values, e.g. the ten or fifteen most important features of female data used in the model.

[0044] In one or more example electronic devices, the one or more processors are configured to determine one or more male biomarkers based on one or more of the male data, the female data, and the pregnancy loss data, e.g. the male data and / or the female data. In one or more example electronic devices, to provide an output comprises to provide a male output associated with, such as representative or indicative of, the one or more male biomarkers.

[0045] The one or more male biomarkers may be indicative of one or more features of the male data influencing the first risk score. For example, to determine the one or more male biomarkers may be based on a SHAP representation of the male data, e.g. input to the machine learning model. To determine one or more male biomarkers may comprise to determine, for each feature or input to the model or for a subset of features, a value, e.g. a SHAP value, indicative of feature importance in the model, optionally followed by a ranking of the values. In other words, to determine one or more male biomarkers may comprise to identify the most important features that have led to the first risk score.

[0046] To provide a male output may comprise outputting a SHAP representation of the male data input to the ML model. In other words, male biomarker(s) may comprise a male data identifier indicative of the male data, such as paternal age, paternal BMI, paternal diastolic blood pressure, paternal pulse, and a value, such as a SHAP value, indicative of feature / data importance in the model. The male output may comprise a subset of SHAP values, e.g. the ten or fifteen most important features of male data used in the model.

[0047] Male data also denoted paternal data are data indicative and / or associated with the male or father of the fetus.

[0048] In one or more example electronic devices, the male data comprises male blood data also denoted MBD of the male. In one or more example electronic devices, to determine the one or more risk scores comprises to determine the one or more risk scores, such as the first risk score, based on the male blood data.

[0049] The male blood data may comprise or be indicative of one or more of Whole Genome Sequencing (WGS), proteomics, metabolomics immunological profile, metabolic profile, inflammatory markers, blood coagulation status, and RNA sequencing. In one or more example electronic devices, the male data comprises male biosample data also denoted MBSD of the male. The male biosample data may comprise one or more of semen data associated with a semen sample, rectal data associated with a rectal sample, and urine data associated with a urine sample. In one or more example electronic devices, to determine the one or more risk scores comprises to determine the one or more risk scores, such as the first risk score, based on the male biosample data, such as one or more of the semen data, the rectal data, and the urine data.

[0050] The semen data may comprise one or more semen parameters, such as sperm count, indicative of semen sample. The semen data of the male biosample data may comprise or be indicative of one or more of DNA fragmentation analysis e.g. by one or more of SDI, COMET and TUNEL, morphology, seminal plasma metabolomics, and immunology.

[0051] The rectal data of the male biosample data may comprise one or more rectal parameters indicative of rectal sample. The rectal data of the male biosample data may comprise or be indicative of Gut microbiome, such as 16S rRNA sequencing and / or shotgunsequencing.

[0052] The urine data of the male biosample data may comprise or be indicative of one or more of endocrine disrupters, cannabinoids, illegal drugs, and cotinine.

[0053] The male data may comprise one or more of age, Body Mass Index (BMI), weight, height, hip-waist ratio, pulse and blood pressure (systolic and / or diastolic) of the male.

[0054] The male data may comprise health data, such as one or more health parameters indicative of male health. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a first health parameter, indicative of whether the male has or has been diagnosed with hypospadias. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a second health parameter, indicative of whether the male has or has been diagnosed with testicular cancer. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a third health parameter, indicative of whether the male has or has been diagnosed with hernia, such as inguinal hernia. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a fourth health parameter, indicative of whether the male has or has been diagnosed with varicocele. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a fifth health parameter, indicative of whether the male has or has been diagnosed with Appendicitis. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a sixth health parameter, indicative of whether the male has or has been diagnosed with genital infection. In one or more examples, the health data of the male data may comprise a health parameter, e.g. a seventh health parameter, indicative of whether the male has had and / or the type of testicular surgery.

[0055] The male data may comprise information or data on one or more of current medications, reproductive history, lifestyle, physical health, and mental health of the male. For example, the male data may comprise male questionnaire data also denoted MQD comprising or indicative of one or more of general information, reproductive history, health, medication and dietary supplements, communication, relationship, staff satisfaction, well-being, sexuality, health behavior, family, childhood, sociodemographic factors, and work. The male questionnaire data may comprise questionnaire data from different times, such as first questionnaire data from questionnaire answers at a first time and / or second questionnaire data from questionnaire answers at a second time. The questionnaire data may comprise third questionnaire data from questionnaire answers at a third time. The questionnaire data of the male data may comprise a parameter indicative of alcohol intake. The questionnaire data of the male data may comprise a parameter indicative of whether the male smokes. Notably, one or more parameters of the questionnaire data may be determined, e.g. via blood and / or urine samples.

[0056] Female data also denoted maternal data are data indicative and / or associated with the female or mother of the fetus.

[0057] In one or more example electronic devices, the female data comprises female blood data also denoted FBD of the female, e.g first female blood data of a first female blood sample taken at a first time and / or second female blood data of a second female blood sample taken at a second time. The first time may be less than 48 hours, such as less than 24 hours, after pregnancy loss. The second time may be larger than 1 week after pregnancy loss, such as in the range from 2 weeks to 8 weeks, e.g. 4 weeks to 6 weeks, after pregnancy loss. In one or more example electronic devices, to determine the one or more risk scores comprises to determine the one or more risk scores, such as the first risk score, based on the female blood data, such as the first female blood data and / or the second female blood data.

[0058] The female blood data may comprise or be indicative of one or more of Whole Genome Sequencing (WGS), Cell-Free Fetal DNA (cffDNA), proteomics, metabolomics immunological profile, metabolic profile, inflammatory markers, blood coagulation status, and RNA sequencing.

[0059] The first female blood sample may be taken at a first time within a first time period, e.g. within 48 hours or within 24 hours of pregnancy loss.

[0060] The second female blood sample may be taken at a second time within a second time period, e.g. in the range from 2 weeks to 8 weeks after pregnancy loss. In one or more examples, the second female blood sample may be taken 4 weeks to 6 weeks after pregnancy loss.

[0061] In one or more example electronic devices, the female data comprises female biosample data also denoted FBSD of the female. The female biosample data FBSD may comprise one or more of vaginal data associated with a vaginal (microbiome) sample, rectal data associated with a rectal sample, and urine data associated with a urine sample. In one or more example electronic devices, to determine the one or more risk scores comprises to determine the one or more risk scores, such as the first risk score, based on the female biosample data, such as one or more of the vaginal data, the rectal data, and the urine data of the female biosample data.

[0062] The vaginal data may comprise one or more vaginal parameters indicative of vaginal sample. The vaginal data of the female biosample data may comprise or be indicative of vaginal microbiome, such as 16S rRNA sequencing and / or shotgun-sequencing and / or vaginal immunology.

[0063] The rectal data of the female data may comprise one or more rectal parameters indicative of rectal sample. The rectal data of the female biosample data may comprise or be indicative of Gut microbiome, such as 16S rRNA sequencing and / or shotgun-sequencing and / or rectal immunology.

[0064] The urine data of the female biosample data may comprise or be indicative of one or more of endocrine disrupters, infections, cannabinoids, illegal drugs, and cotinine.

[0065] In one or more example electronic devices, the female data comprises ultrasound image data of an ultrasound scanning of the uterus of the female. To determine the one or more risk scores may comprise to determine the one or more risk scores, such as the first risk score, based on the ultrasound image data. The female data may comprise one or more of type of pregnancy loss, selected evacuation treatment, and if and / or which pregnancy complications occurred. A type of pregnancy loss may be selected from one or more of spontaneous complete miscarriage, spontaneous incomplete miscarriage, missed abortion, and anembryonic pregnancy.

[0066] The female data may comprise one or more of age, Body Mass Index (BMI), weight, height, hip-waist ratio, pulse and blood pressure (systolic and / or diastolic) of the female.

[0067] The female data may comprise health data, such as one or more health parameters indicative of female health. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a first health parameter, indicative of whether the female has, has been operated for, or has been diagnosed with endometriosis. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a second health parameter, indicative of whether the female has or has been diagnosed with fibrom, such as uterine fibroms. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a third health parameter, indicative of whether the female has, has been operated for, or has been diagnosed with hernia. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a fourth health parameter, indicative of whether the female has, has been operated for, or has been diagnosed with ovarian cyst. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a fifth health parameter, indicative of whether the female has, has been operated for, or has been diagnosed with Appendicitis. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a sixth health parameter, indicative of whether the female has or has been diagnosed with genital infection. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a seventh health parameter, indicative of whether the female has, has been operated for, or has been diagnosed with Ectopic pregancy. In one or more examples, the health data of the female data may comprise a health parameter, e.g. an eighth health parameter, indicative of whether the female has, has been operated for, or has been diagnosed with Polyscystiuc Ovarian Syndrome. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a ninth health parameter, indicative of whether the female has had a Caesarean section. In one or more examples, the health data of the female data may comprise a health parameter, e.g. a tenth health parameter, indicative of whether the female has had vaginal bleeding during pregnancy. The health data may be obtained as questionnaire data via a questionnaire and / or via a patient record. The female data may comprise information or data on one or more of current medications, reproductive history, lifestyle, physical health, and mental health of the female. For example, the female data may comprise female questionnaire data also denoted FQD comprising or indicative of one or more of general information, reproductive history, health, medication and dietary supplements, communication, relationship, staff satisfaction, well-being, sexuality, health behavior, family, childhood, sociodemographic factors, and work. The questionnaire data of the female data may comprise questionnaire data from different times, such first questionnaire data from questionnaire answers at a first time and / or second questionnaire data from questionnaire answers at a second time. The questionnaire data may comprise third questionnaire data from questionnaire answers at a third time. The questionnaire data of the female data may comprise a parameter indicative of alcohol intake during pregnancy. The questionnaire data of the female data may comprise a parameter indicative of smoking during pregnancy. The female data, such as questionnaire data, may comprise a partner parameter indicative of the number of sexual partners.

[0068] The female data, such as questionnaire data, may comprise a first dietary parameter, e.g. indicative of whether the female has taken vitamin(s) prior to and / or during the pregnancy. For example the first dietary parameter may be a vector or matrix and may be indicative of type(s) and / or amount(s) of vitamin supplements, such as vitamin D and / or vitamin E, taken prior to and / or during pregnancy.

[0069] The female data, such as questionnaire data, may comprise a second dietary parameter, e.g. indicative of whether the female has taken fish oil prior to and / or during the pregnancy. For example, the second dietary parameter may be a vector or matrix and may be indicative of type(s) and / or amount(s) of fish oil taken prior to and / or during pregnancy.

[0070] The pregnancy loss data may be associated with or indicative of one or more of a fetus, fetal tissue, pregnancy tissue, and product of conception and / or tissue thereof.

[0071] The pregnancy loss data may comprise one or more parameters indicative of one or more of method or mode of conception, such as natural, IVF, or insemination, gestational age(s), e.g. based on last menstrual period also denoted GA_LM and / or findings from diagnostic ultrasound (e.g. multiple gestations) also denoted GA_UL, and phenotype. The pregnancy loss data may comprise one or more parameters indicative whether a gestational age estimated from crown-rump length could have been obtained and / or was not obtainable.

[0072] The pregnancy loss data may comprise one or more parameters indicative of whether donor sperm and / or donor oocyte was used for conception of the fetus.

[0073] The pregnancy loss data may comprise one or more parameters indicative of whether donor sperm and / or donor oocyte was used for conception of the fetus.

[0074] The pregnancy loss data may comprise one or more of type of pregnancy loss, selected evacuation treatment, and if and / or which complications occurred. A type of pregnancy loss may be selected from one or more of spontaneous complete miscarriage, spontaneous incomplete miscarriage, missed abortion, and anembryonic pregnancy.

[0075] The pregnancy loss data may comprise fetal biosample data of the fetus. The fetal biosample data may be associated with or be determined from one or more of pregnancy tissue, atretic embryos, and oocytes. The fetal biosample data may comprise one or more of Whole Genome Sequencing (WGS) and Single-cell shallow sequencing.

[0076] In one or more example electronic devices, to determine the one or more risk scores, such as the first risk score, comprises to apply a machine learning model, e.g. to one or more of the male data MD, the female data FD, and the pregnancy loss data PLD. In other words, male data, female data, and optionally pregnancy loss data may be fed to a machine learning model providing as output one or more risk scores, e.g. including first risk score. The male data, such as one or more of male blood data, male biosample data, and health data of the male as described herein, may be fed to the machine learning model for provision of one or more risk scores, e.g. including first risk score(s), as output. The female data, such as one or more of female blood data, female biosample data, and health data of the female as described herein, may be fed to the machine learning model for provision of one or more risk scores, e.g. including first risk score(s), as output.

[0077] In one or more example electronic devices, the first risk score also denoted RS_1 comprises or is a first primary risk score also denoted RS_1_1, e.g. indicative of a possibility of a euploid loss. The first risk score may be a single value, e.g. indicative of a possibility of a euploid loss or a indicative of a possibility of an aneuploid loss. In one or more example electronic devices, the first risk score RS_1 comprises a first secondary risk score also denoted RS_1_2, e.g. indicative of a possibility of an aneuploid loss. In other words, the first risk score may comprise a plurality of first risk scores including a first primary risk score indicative of a possibility of a euploid loss and a first secondary risk score indicative of a possibility of an aneuploid loss.

[0078] The one or more risk scores may comprise a second risk score also denoted RS_2 and / or a third risk score also denoted RS_3.

[0079] In other words, to determine one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, may comprise to determine a second risk score and / or a third risk score.

[0080] To provide an output may comprise to provide an output including a second output associated with the second risk score. To provide an output may comprise to provide an output including a third output associated with the third risk score.

[0081] A computer-implemented method for pregnancy loss classification is disclosed. The method comprises obtaining male data associated with a male. The method comprises obtaining female data associated with a female. The method comprises obtaining pregnancy loss data associated with a fetus and / or fetal tissue. The method comprises determining one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status. The method comprises providing an output including a first output associated with the one or more risk scores.

[0082] In one or more example computer-implemented methods, the method comprises determining one or more female biomarkers based on the female data and / or based on the machine learning model. In one or more example methods for pregnancy loss classification, providing an output comprises providing a female output associated with the one or more female biomarkers.

[0083] In one or more example computer-implemented methods, the method comprises determining one or more male biomarkers based on the male data and / or based on the machine learning model. In one or more example methods for pregnancy loss classification, providing an output comprises providing a male output associated with the one or more male biomarkers.

[0084] In one or more example computer-implemented methods, the male data MD comprises male blood data MBD of the male. In one or more example methods for pregnancy loss classification, determining the one or more risk scores comprises determining the one or more risk scores based on the male blood data.

[0085] In one or more example computer-implemented methods, the male data MD comprises male biosample data MBSD of the male. In one or more example methods for pregnancy loss classification, determining the one or more risk scores comprises determining the one or more risk scores based on the male biosample data.

[0086] In one or more example computer-implemented methods, the female data FD comprises female blood data FBD of the female. In one or more example methods for pregnancy loss classification, determining the one or more risk scores comprises determining the one or more risk scores based on the female blood data.

[0087] In one or more example computer-implemented methods, the female data FD comprises female biosample data FBSD of the female. In one or more example methods for pregnancy loss classification, determining the one or more risk scores comprises determining the one or more risk scores based on the female biosample data.

[0088] In one or more example computer-implemented methods, determining the one or more risk scores comprises applying a machine learning model, such as a neural network.

[0089] In one or more example computer-implemented methods, the first risk score comprises a first primary risk score indicative of a possibility of a euploid loss.

[0090] In one or more example computer-implemented methods, the first risk score comprises a first secondary risk score indicative of a possibility of an aneuploid loss.

[0091] A computer-implemented method for training a neural network to process as inputs one or more, such as two or all of, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue, and to provide as output one or more risk scores associated with pregnancy loss is disclosed. The method comprises obtaining, using at least one processor, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue. The method comprises performing, using the at least one processor, a training, wherein performing the training comprises generating, using the at least one processor and a machine-learning model, risk data based on the male data, the female data, and the pregnancy loss data. Performing the training comprises obtaining, using the at least one processor, training data; determining, using the at least one processor and one or more loss functions, one or more loss parameters based on the male data, the female data, the pregnancy loss data, and the training data; and training, using the at least one processor, the machine learning model based on the one or more loss parameters.

[0092] It is to be understood that a description of a feature in relation to the electronic device is also applicable to the corresponding feature in the method(s) of operating an electronic device as disclosed herein.

[0093] Fig. 1 illustrates a system implementing the electronic device and / or methods according to the present disclosure. The system 2 comprises one or more databases 4 storing clinical data including one or more of male data MD, female data FD, and pregnancy loss data PLD. The system 2 comprises an electronic device 6 for pregnancy loss classification, the electronic device comprising an interface 8, one or more processors including processor 10, and a memory 12, wherein the one of more processors are configured to obtain, via the interface 8, the male data MD associated with a male; obtain, via the interface 8, female data FD associated with a female; and obtain, via the interface 8, pregnancy loss data PLD associated with a fetus and / or fetal tissue. The one or more databases 4 communicates with the electronic device 6 over network. The electronic device 6 may implement database 4A in memory 12.

[0094] The one of more processors are configured to determine one or more risk scores based on one or more of the male data MD, the female data FD, and the pregnancy loss data PLD. The one or more risk scores are associated with pregnancy loss and includes a first risk score RS_1 indicative of a fetal ploidy status. The one of more processors are configured to provide, via the interface 8, an output including a first output associated with one or more risk scores, such as the first risk score RS_1 . The interface 8 of the electronic device optionally comprises a display 14, wherein to provide an output comprises to display a first user interface element 30 as a first output associated or representing the first risk score (RS_1=0.8 in the illustrated example). For example, to provide an output may comprise to display a first primary user interface element as a first primary output associated or representing the first primary risk score. For example, to provide an output may comprise to display a first secondary user interface element as a first secondary output associated or representing the first secondary risk score.

[0095] The electronic device 6 may be configured to perform any of the methods disclosed herein, such as the method in any of Fig. 2 and / or Fig. 3. In other words, the electronic device 6 is configured for analysis, classification, monitoring, and / or prediction of pregnancy loss and / or pregnancy complications.

[0096] The processor circuitry 302 is optionally configured to perform any of the operations disclosed in Figs. 2-3 (such as any one or more of: S102, S104, S106, S108, S110, S112, S112A, S112B, S112C, S112D, S112E, 114, S114A, S114B, S202, S204, S204A, S204B, S204C, S204D). The operations of the electronic device 6 may be embodied in the form of executable logic routines (e.g., lines of code, software programs, etc.) that are stored on a non-transitory computer readable medium (e.g., the memory 12) and are executed by the one or more processors / processor 10.

[0097] Furthermore, the operations of the electronic device 6 may be considered a method that the electronic device 6 is configured to carry out. Also, while the described functions and operations may be implemented in software, such functionality may as well be carried out via dedicated hardware or firmware, or some combination of hardware, firmware and / or software.

[0098] The memory 12 may be or comprise one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, a random access memory (RAM), or other suitable device. In a typical arrangement, the memory 12 may include a non-volatile memory for long term data storage and a volatile memory that functions as system memory for the processor 10. The memory 12 may exchange data with the processor 10 over a data bus. Control lines and an address bus between the memory 12 and the processor 10 also may be present (not shown in Fig. 1 ). The memory 12 is considered a non-transitory computer readable medium.

[0099] The memory 12 may be configured to store one or more of male data, such as male blood data and / or male biosample data, female data, such as female blood data and / or, female biosample data, pregnancy loss data, one or more risk scores, e.g. including a first risk score, an output, e.g. including a first output, a male output, a female output, one or more female biomarkers, one or more male biomarkers, machine learning model, first primary risk score, first secondary risk score, risk data, training data and / or one or more loss parameters in a part of the memory 12.

[0100] Fig. 2 shows a flow chart of an example computer-implemented method for pregnancy loss classification. The method 100 comprises obtaining S102 male data associated with a male. The method comprises obtaining S104 female data associated with a female. The method optionally comprises obtaining S106 pregnancy loss data associated with a fetus and / or fetal tissue. S102, S104, and S106 may be collectively referred to obtaining S101 clinical data. In one or more examples, the method 100 comprises determining S108 one or more female biomarkers, e.g. based on the female data. The one or more female biomarkers may comprise a first female biomarker and / or a second female biomarker. In one or more examples, the method 100 comprises determining S110 one or more male biomarkers, e.g. based on the male data. The one or more male biomarkers may comprise a first male biomarker and / or a second male biomarker. The method comprises determining S112 one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and optionally including a first risk score indicative of a fetal ploidy status. The method comprises providing S114 an output including a first output associated with the one or more risk scores, such as the first risk score. Providing S114 an output comprises providing S114A a first output associated with the first risk score, e.g. by displaying S114D a first user interface element as a first output representing the first risk score.

[0101] Fig. 3 shows a flow chart of an example computer-implemented method for pregnancy loss classification. The method 100A comprises obtaining S102 male data associated with a male. The method comprises obtaining S104 female data associated with a female. The method comprises obtaining S106 pregnancy loss data associated with a fetus and / or fetal tissue. The method comprises determining S112 one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status. The method comprises providing S114 an output including providing S114A a first output associated with the first risk score.

[0102] Fig. 4 shows a flow chart of an example computer-implemented method for training a machine learning model, such as a neural network, to process as inputs male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue, and for providing as output one or more risk scores associated with pregnancy loss. The machine learning model / neural network may be for providing as output one or more male biomarkers and / or one or more female biomarkers. The method 200 comprises obtaining S202, using at least one processor, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue. The method comprises performing S204, using the at least one processor, a training. Performing S204 the training comprises generating S204A, using the at least one processor and a machine-learning model, such as a neural network, risk data based on the male data, the female data, and the pregnancy loss data. Performing S204 the training comprises obtaining S204B, using the at least one processor, training data. Performing S204 the training comprises determining S204C, using the at least one processor and one or more loss functions, one or more loss parameters based on the male data, the female data, the pregnancy loss data, and the training data. Performing S204 the training comprises training S204D, using the at least one processor, the machine learning model / neural network based on the one or more loss parameters.

[0103] Fig. 5 illustrates an example implementation of a machine learning model according to the present disclosure. The machine learning model 10A takes as input male data, female data, and pregnancy loss data PLD, the male data MD optionally including one or more of MBD, MBSD, and MQD, and the female data FD optionally including one or more of FBD, FBSD, and FQD. The machine learning model 10A provides as output a first risk score RS_1 , such as one or both of RS_1_1 indicative of a possibility of a euploid loss and RS_1_2 indicative of a possibility of an aneuploid loss. The machine learning model 10A optionally provides as output one or both of male biomarkers MB_1-MB_N and female biomarkers FB_1-FB_M, where N is the number of male biomarkers and M is the number of female biomarkers.

[0104] EXAMPLE

[0105] In an example, the machine-learning model is trained on a first cohort also denoted Cohort 1 of couples having experienced pregnancy loss, where the ploidy status of the fetal was determined via cffDNA-based test results in which the maternal blood was drawn while the pregnancy tissue was still in situ or shortly after, the fetal ploidy status was evaluated by the principles of Non-lnvasive Prenatal Testing (NIPT) by next- generation sequencing. The results were given as euploidy (46, XX or 46, XY) or aneuploidy (monosomy, trisomy, or multiple trisomies).

[0106] Training data of the first cohort included male (paternal) data and female (maternal) data including BMI, blood pressure, health data including conditions such as endometriosis, PCOS, and testicular cancer, and behavioral factors (questionnaire data), such as the number of previous sexual partners, risk-prone behaviors, vitamin intake (D and / or E). Further, pregnancy loss data were used in the training. A comprehensive list of these parameters / data is provided in Table 1 below. The data were obtained via electronic health record (EHR) and questionnaire as indicated.

[0107] Table 1. Parameters / data used for training of machine learning model for provision of first risk score.

[0108] The machine learning model may be implemented as a gradient boosting model, such as a lightGBM. A gradient boosting model is fast, natively handles missing values, and encodes categorical variables using an algorithm superior to one-hot encoding. To find the best set of hyperparameters, each hyperparameter setting was evaluated in five-fold cross-validation process, repeated three times. The optimal model was selected based on cross-validated negative log-likelihood. Hyperparameters were optimized using Optuna software, run for 500 trials with default settings.

[0109] The outcome, euploid versus aneuploid pregnancy loss, was modelled using LightGBM (Light Gradient Boosting Machine). LightGBM employs a gradient boosting framework to construct an ensemble of decision trees, aiming to minimize a differentiable loss function L(y,F(x)), where y is the true label and F(x) is the predicted probability. The loss function is defined as the binary cross entropy,

[0110] L(y,F (x)) = — [y • log(p) + (1 - y) • log(l - p)] LightGBM was chosen as it is fast, natively handles missing values, and encodes categorical variables using an algorithm superior to one-hot encoding.

[0111] LightGBM is a complex model with many potential hyperparameters to tune. For this application, the model was tuned to the maximum depth, minimum number of child samples, learning rate, number of estimators, fraction of examples subsampled, and fraction of features subsampled.

[0112] To identify the best set of hyperparameters, each combination of hyperparameters was evaluated in five-fold cross-validation process, repeated three times. To avoid using a grid or random search, the hyperparameter optimization framework Optuna was employed. Optuna implements the Tree-structured Parzen Estimator (TPE), which attempts to find the best set of hyperparameters in a given search space by modeling the objective function as a probabilistic distribution and subsequently sampling new candidate hyperparameters based on this model, thus optimizing both exploration and exploitation strategies.

[0113] The optimal model was selected based on cross-validated negative cross entropy. Hyperparameters were optimized using Optuna software, run for 500 trials with default settings. The cross-validated Area Under the Receiver Operating Characteristic (AUC- ROC) and Area Under the Precision-Recall Curve (AUC-PRC) were used as a measure of internal validation.

[0114] Feature importance was dissected using Shapley Additive Explanations (SHAP). The SHAP algorithm was applied to Cohort 2 and Cohort 3 to ascertain the contribution of each feature to the risk of euploid pregnancy loss. Furthermore, the SHAP scores / values were compared to the training data set. It's imperative to note that SHAP values are computed for each feature for each patient, thus furnishing one or more individual biomarkers and an individualized estimate of contribution to the resultant first risk score.

[0115] The analysis was performed using Snakemake, Optuna, Scikit-learn, and Python interfaces for LightGBM and SHAP.

[0116] The model displayed acceptable generalizability. The AUC-ROC scores from Cohort 2 (0.69; 0.62-0.76 95% Cl) and Cohort 3 (0.66; 0.59-0.74 95% Cl) were similar to the results from cross-validation (Fig. 6). The AUC-PRC was highest at Cohort 2 (0.73; 0.64-0.80 95% Cl) compared to Cohort 3 (0.66; 0.56-0.76 95% Cl), although both were comparable to the cross-validation (Fig. 7). Visual inspection of calibration plots (Fig. 8 and 9) indicated that the predicted probability was well aligned with the observed prevalence in both Cohort 2 and Cohort 3.

[0117] Model development was done using the data collected from the first cohort. External model validation was performed using data from a second cohort also denoted Cohort 2 and a third cohort also denoted Cohort 3 from two separate inclusion sites, which were not involved in the development of the model. Cross-validated Area Under the Receiver Operating Characteristic (AUC-ROC) and Area Under the Precision-Recall Curve (AUC- PRC) shown in Fig. 6 and Fig. 7 are used for validation and confidence intervals were generated using 1 ,000 bootstrap samples.

[0118] Feature importance was dissected using Shapley Additive Explanations (SHAP). The SHAP algorithm was applied to the two external validation datasets to ascertain the contribution of each feature to the risk of euploid pregnancy loss. Furthermore, we compare the SHAP scores to the data of the first Cohort. The analysis was performed using Snakemake, Optuna, Scikit-learn, and Python interfaces for LightGBM and SHAP.

[0119] SHAP ANALYSIS

[0120] To dissect the decision-making process of the LightGBM model, SHAP (SHapley Additive exPlanations) values, a state-of-the-art approach to model interpretation, were applied, e.g. to determine one or more biomarkers for a respective couple, see Figs. 10-14.

[0121] Referring to Fig. 10 it is seen that summarized across all features, maternal age was the most important. However, the gestational age calculated from last menstruation was nearly as important. Strikingly, Vitamin D and Vitamin E supplementation was also important. The paternal BMI, age, and blood pressure all affected the likelihood of an euploid pregnancy loss. Maternal vitamin D supplement had a magnitude similar to that of paternal age.

[0122] Furthermore, the SHAP analysis can also be used to inform which variables led to the probability of an euploid loss. In other words, the SHAP analysis can also be used to identify one or more (individual) biomarkers for pregnancy loss. We showcase two examples, in which we have selected the female with the highest and lowest predicted probability of an euploid loss.

[0123] In the first case, the predicted probability was 80%. The major contributing factors were gestational age calculated from last menstruation, maternal age, gestational age estimated from ultrasound (Fig. 11). Conversely, in the second case, the predicted probability was 15.3%. The major contributing factors were maternal age, gestational age calculated from last menstruation, and Vitamin E supplement (Fig. 12).

[0124] The presented model or algorithm is also capable of providing individual explanations. This is exemplified by contrasting two individuals that had the same predicted probability by the model, namely 71 %, and both had an euploid pregnancy loss. The ranking of the most important factors is very different between the two cases, see Fig. 13 and 14.

[0125] The integrative SHAP analysis provides a way to give feedback and describing to clinicians and patients which factors contributed to a low or high prediction of euploid loss.

[0126] Fig. 6 illustrates an AUC-ROC curve and Fig. 7 illustrates an AUC-PRC curve for an example model trained on male data and female data of Cohort 1 . The female data used for training included data / parameters indicative of age, BMI, systolic blood pressure, diastolic blood pressure, and pulse. Further, female data used for training the model included data / parameters indicative of one or more of vaginal bleeding, pain at inclusion, Polycystic Ovary Syndrome, Endometriosis, Uterine fibroms, prior Caesarean section, prior operation for Ectopic Pregnancy, prior operation for Ovarian cysts, prior operation for Hernia, prior operation for Appendicitis, other operations, age at menarche, number of sexual partners, pain during menstruation, risk taking, Genital Infections, Vitamin D supplement during pregnancy, Vitamin E supplement during pregnancy, and smoking during pregnancy. Further female data used for training included data / parameters indicative of one or more of para, number of prior losses, mode of conception, donor sperm, diagnosis (Missed abortion, bligted ovum, spontanous ongoing pregnancy loss). Pregnancy loss data, e.g. including gestational age calculated from last menstruation and / or gestational age calculated from crown-rump length was also used in the training.

[0127] The male data used for training included data / parameters indicative of age, BMI, systolic blood pressure, diastolic blood pressure, and pulse. Further, male data used for training the model included data / parameters indicative of one or more of smoking, alcohol, Genital Infections, Appendicitis, Inguinal hernia, and Testicular surgery.

[0128] Fig. 6 shows presents the Receiver Operating Characteristic (ROC) curves for the two external validation cohort, Cohort 2 and Cohort 3, in evaluating the performance for classification of euploid pregnancy loss. The ROC curve plots the True Positive Rate (Sensitivity) against the False Positive Rate (1 -Specificity) at various classification thresholds. The Area Under the Curve (AUC) values are computed using the trapezoidal rule. The diagonal dashed line represents a random classifier (AUC = 0.5) serving as a baseline for comparison. Both cohorts indicate that the model generalizes well and aligns well with the cross-validated AUC-ROC of 0.69.

[0129] Fig. 7 shows the Precision-Recall Curves (PRC) for the two validation cohorts — Cohort 2 and Cohort 3 — aimed at classifying euploid pregnancy loss. Precision-Recall Curves graphically represent the trade-off between Precision (Positive Predictive Value) and Recall (Sensitivity) at different classification thresholds. A horizontal dashed line represents the no-skill classifier, characterized by a precision equal to the proportion of positive samples in the dataset. The AUC-PRC values are calculated using appropriate numerical methods such as the trapezoidal rule. The auPRC is, in both cases, superior to a no-skill classifier (i.e. > 0.49), and the values for Cohort 2 (auPRC = 0.73) and Cohort 3 (auPRC = 0.65) is in the same order as the cross-validated value from Cohort 1 (auPRC = 0.68).

[0130] Fig. 8 shows a calibration curve for Cohort 2 to assess the predictive performance of the model in estimating the probability of euploid pregnancy loss. Calibrated Probabilities (Solid Line): The solid line represents the model's predicted probabilities plotted against the observed frequencies. A perfectly calibrated model would follow the grey 45-degree diagonal line. This serves as a reference line for assessing model performance. Calibration-in-the-Large and Calibration Slope: The plot includes annotations for calibration-in-the-large (Intercept = 0.065) and the calibration slope (Slope = 1.175), indicating that the model is well calibrated. Deviations from a calibration-in-the-large of 0 or a calibration slope of 1 indicate some level of miscalibration. The average prediction error, Eavg, was 2.9%.

[0131] Fig. 9 depicts a calibration curve for to assess the predictive performance of the model in estimating the probability of euploid pregnancy loss in Cohort 3. The description follows the one from Figure 8. The model was well calibrated, as indicated by the calibration-in- the-large (Intercept = -0.056) and calibration slope (Slope = 1.026). The average prediction error, Eavg, was 3.1 %.

[0132] Fig. 10 presents a SHAP (SHapley Additive exPlanations) bar plot, identifying the 15 most important features in the predictive model according to their absolute mean SHAP values. The SHAP values, which are on the log-odds scale, provide a measure of the impact of each feature on the model output, considering the magnitude feature's effect. Features are ranked from top to bottom in descending order based on their absolute mean SHAP values. Fig. 11 shows a Shapley Additive Explanations (SHAP) bar plot for the investigated individual with the highest predicted probability of an euploid loss. The first (primary ) risk score indicative of probability of euploid loss output or estimated from the model was 0.80, and the individual did have an euploid pregnancy loss. The values on the left side of the figure are the actual individual values. The major contributing factors to the prediction were fetal gestational age calculated from date of last menstruation, maternal age, and the lack of a fetal gestational age estimated from ultrasound (represented by a nan).

[0133] Fig. 12 shows a Shapley Additive Explanations (SHAP) bar plot for the investigated individual with the lowest predicted probability of an euploid loss. The first (primary) risk score indicative of probability of euploid loss output or estimated from the model was 15.3%, and the individual did not have an euploid pregnancy loss. The major contributing factors to the prediction was maternal age, fetal gestational age calculated from date of last menstruation, and vitamin E supplement. In this case, Vitamin E supplement during pregnancy lowered the probability of an euploid pregnancy loss.

[0134] Fig. 13 shows a Shapley Additive Explanations (SHAP) bar plot for an investigated individual. The individual had a first (primary) risk score indicative of probability of euploid loss output or estimated from the model of 0.71 , and the individual did have an euploid pregnancy loss. The most important factors were paternal age, fetal gestational age calculated from last menstruation, and maternal age. The maternal systolic blood pressure, ranking #7, had a SHAP value of 0.06.

[0135] Fig. 14 shows a Shapley Additive Explanations (SHAP) bar plot for an investigated individual. The individual had a first (primary) risk score indicative of probability of euploid loss output or estimated from the model of 0.71 , and the individual did have an euploid pregnancy loss. This is the same outcome as the case presented in Fig. 13. However, in contrast to Fig. 13, the most important factors in this case were fetal gestational age calculated from last menstruation, fetal age estimated from ultrasound, and maternal age. Moreover, the maternal systolic blood pressure had a SHAP value nearly twice the magnitude of the case presented in Fig. 13.

[0136] The use of the terms “first”, “second”, “third” and “fourth”, “primary”, “secondary”, “tertiary” etc. does not imply any particular order, but are included to identify individual elements. Moreover, the use of the terms “first”, “second”, “third” and “fourth”, “primary”, “secondary”, “tertiary” etc. does not denote any order or importance, but rather the terms “first”, “second”, “third” and “fourth”, “primary”, “secondary”, “tertiary” etc. are used to distinguish one element from another. Note that the words “first”, “second”, “third” and “fourth”, “primary”, “secondary”, “tertiary” etc. are used here and elsewhere for labelling purposes only and are not intended to denote any specific spatial or temporal ordering unless otherwise specified. Furthermore, the labelling of a first element does not imply the presence of a second element and vice versa.

[0137] It may be appreciated that the figures comprise some circuitries, blocks, or operations which are illustrated with a solid line and some circuitries, components, features, or operations which are illustrated with a dashed line. Circuitries or operations which are comprised in a solid line are circuitries, components, features or operations which are comprised in the broadest example. Circuitries, components, features, or operations which are comprised in a dashed line are optional examples which may be comprised in, or a part of, or are further circuitries, components, features, or operations which may be taken in addition to circuitries, components, features, or operations of the solid line examples. It should be appreciated that these operations need not be performed in order presented. Furthermore, it should be appreciated that not all the operations need to be performed. The example operations may be performed in any order and in any combination. It should be appreciated that these operations need not be performed in order presented. Circuitries, components, features, or operations which are comprised in a dashed line may be considered optional.

[0138] Other operations that are not described herein can be incorporated in the example operations. For example, one or more additional operations can be performed before, after, simultaneously, or between any of the described operations.

[0139] Certain features discussed above as separate implementations can also be implemented in combination as a single implementation. Conversely, features described as a single implementation can also be implemented in multiple implementations separately or in any suitable sub-combination. Moreover, although features may be described above as acting in certain combinations, one or more features from a claimed combination can, in some cases, be excised from the combination, and the combination may be claimed as any subcombination or variation of any sub-combination.

[0140] It is to be noted that the word "comprising", “comprise” or “comprises” does not exclude the presence of other elements or steps than those listed.

[0141] It is to be noted that the words “a”, “an” or “the” preceding an element do not exclude the presence of a plurality of such elements. It should be noted that the use of “or” means “and / or” unless stated otherwise.

[0142] It should be noted that the use of “parameter” also means “weight” in a network based model, and also means amount of training cycles, and also means step-size and also means binning size and stratification schemes for cross validation, and also means data fit, and also means error function, and also means choice of activation function, and also means choice of one or more types of machine learning methods to be applied, and also means any machine learning related choice made by operator and / or optimisation algorithm unless stated otherwise.

[0143] It should further be noted that any reference signs do not limit the scope of the claims, that the examples may be implemented at least in part by means of both hardware and software, and that several “mean”, “unit” or “device” may be represented by the same item of hardware.

[0144] Language of degree used herein, such as the terms “approximately,” “about,” “generally,” and “substantially” as used herein represent a value, amount, or characteristic close to the stated value, amount, or characteristic that still performs a desired function or achieves a desired result. For example, the terms “approximately”, “about”, “generally,” and “substantially” may refer to an amount that is within less than or equal to 10% of, within less than or equal to 5% of, within less than or equal to 1% of, within less than or equal to 0.1 % of, and within less than or equal to 0.01% of the stated amount. If the stated amount is 0 (e.g., none, having no), the above recited ranges can be specific ranges, and not within a particular % of the value. For example, within less than or equal to 10 wt. / vol. % of, within less than or equal to 5 wt. / vol. % of, within less than or equal to 1 wt. / vol. % of, within less than or equal to 0.1 wt. / vol. % of, and within less than or equal to 0.01 wt. / vol. % of the stated amount.

[0145] Although features have been shown and described, it will be understood that they are not intended to limit the claimed disclosure, and it will be made obvious to those skilled in the art that various changes and modifications may be made without departing from the scope of the claimed disclosure. The specification and drawings are, accordingly, to be regarded in an illustrative rather than restrictive sense. The claimed disclosure is intended to cover all alternatives, modifications, and equivalents.

[0146] Examples of the electronic device and related method according to the disclosure are set out in the following items: Item 1 . An electronic device for pregnancy loss classification, the electronic device comprising an interface, one or more processors, and a memory, wherein the one of more processors are configured to: obtain male data associated with a male; obtain female data associated with a female; obtain pregnancy loss data associated with a fetus and / or fetal tissue; determine one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status; and provide an output including a first output associated with the one or more risk scores.

[0147] Item 2. Electronic device according to item 1 , wherein the one or more processors are configured to determine one or more female biomarkers based on the female data, and wherein to provide an output comprises to provide a female output associated with the one or more female biomarkers.

[0148] Item 3. Electronic device according to any one of items 1-2, wherein the one or more processors are configured to determine one or more male biomarkers based on the male data, and wherein to provide an output comprises to provide a male output associated with the one or more male biomarkers.

[0149] Item 4. Electronic device according to any one of items 1-3, wherein the male data comprises male blood data of the male, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the male blood data.

[0150] Item 5. Electronic device according to any one of items 1-4, wherein the male data comprises male biosample data of the male, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the male biosample data.

[0151] Item 6. Electronic device according to item 5, wherein the male biosample data comprises semen data of the male, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the semen data. Item 7. Electronic device according to any one of items 5-6, wherein the male biosample data comprises rectal data, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the rectal data of the male biosample data.

[0152] Item 8. Electronic device according to any one of items 5-7, wherein the male biosample data comprises urine data, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the urine data of the male biosample data.

[0153] Item 9. Electronic device according to any one of items 1-8, wherein the male data comprises male questionnaire data, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the male questionnaire data.

[0154] Item 10. Electronic device according to any one of items 1-9, wherein the female data comprises female blood data of the female, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the female blood data.

[0155] Item 11. Electronic device according to any one of items 1-10, wherein the female data comprises female biosample data of the female, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the female biosample data.

[0156] Item 12. Electronic device according to item 11 , wherein the female biosample data comprises vaginal data of the female, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the vaginal data.

[0157] Item 13. Electronic device according to any one of items 11-12, wherein the female biosample data comprises rectal data, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the rectal data of the female biosample data.

[0158] Item 14. Electronic device according to any one of items 11-13, wherein the female biosample data comprises urine data, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the urine data of the female biosample data. Item 15. Electronic device according to any one of items 1-8, wherein the female data comprises female questionnaire data, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the female questionnaire data.

[0159] Item 16. Electronic device according to any one of items 1-15, wherein to determine the one or more risk scores comprises to apply a machine learning model e.g. to one or more of the male data, the female data, and the pregnancy loss data.

[0160] Item 17. Electronic device according to any one of items 1-16, wherein the first risk score comprises a first primary risk score indicative of a possibility of a euploid loss.

[0161] Item 18. Electronic device according to any one of items 1-17, wherein the first risk score comprises a first secondary risk score indicative of a possibility of an aneuploid loss.

[0162] Item 19. A computer-implemented method for pregnancy loss classification, the method comprising: obtaining male data associated with a male; obtaining female data associated with a female; obtaining pregnancy loss data associated with a fetus and / or fetal tissue; determining one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status; and providing an output including a first output associated with the one or more risk scores.

[0163] Item 20. A computer-implemented method for training a machine learning model, such as a neural network, to process as inputs male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue, and provide as output one or more risk scores associated with pregnancy loss, the method comprising: obtaining, using at least one processor, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue; performing, using the at least one processor, a training comprising: generating, using the at least one processor and a machine-learning model, risk data based on the male data, the female data, and the pregnancy loss data; obtaining, using the at least one processor, training data; determining, using the at least one processor and one or more loss functions, one or more loss parameters based on the male data, the female data, the pregnancy loss data, and the training data; and training, using the at least one processor, the machine learning model based on one or more loss parameters.

[0164] LIST OF REFERENCES

[0165] 2 system

[0166] 4, 4A database

[0167] 6 electronic device

[0168] 7 network

[0169] 8 interface

[0170] 10 processor

[0171] 10A machine learning model I neural network

[0172] 12 memory

[0173] 14 display

[0174] 16 network interface

[0175] 30 first user interface element

[0176] 100 method for pregnancy loss classification

[0177] 5101 obtaining clinical data

[0178] 5102 obtaining male data

[0179] S104 obtaining female data

[0180] S106 obtaining pregnancy loss data

[0181] S108 determining one or more female biomarkers

[0182] S110 determining one or more male biomarkers

[0183] S112 determining one or more risk scores

[0184] S112A determining the one or more risk scores based on the male blood data

[0185] S112B determining the one or more risk scores based on the male biosample data S112C determining the one or more risk scores based on the female blood data

[0186] S112D determining the one or more risk scores based on the female biosample data

[0187] S112E applying a machine learning model

[0188] S114 providing an output

[0189] S114A providing a first output associated with the first risk score

[0190] S114B providing a female output associated with one or more female biomarkers

[0191] S114C providing a male output associated with one or more male biomarkers

[0192] 200 method for training a neural network

[0193] S202 obtaining male data, female data, and pregnancy loss data

[0194] S204 performing a training

[0195] S204A generating risk data based on the male data, the female data, and the pregnancy loss data

[0196] S204B obtaining training data

[0197] S204C determining one or more loss parameters

[0198] S204D training the neural network

[0199] FD female data

[0200] FBD female blood data

[0201] FBSD female biosample data

[0202] FQD female questionnaire data

[0203] MD male data

[0204] MBD male blood data

[0205] MBSD male biosample data MQD male questionnaire data

[0206] PLD pregnancy loss data

Claims

CLAIMS1 . An electronic device for pregnancy loss classification, the electronic device comprising an interface, one or more processors, and a memory, wherein the one of more processors are configured to: obtain male data associated with a male; obtain female data associated with a female; obtain pregnancy loss data associated with a fetus and / or fetal tissue; determine one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status; and provide an output including a first output associated with the one or more risk scores.

2. Electronic device according to claim 1 , wherein the one or more processors are configured to determine one or more female biomarkers based on the female data, and wherein to provide an output comprises to provide a female output associated with the one or more female biomarkers.

3. Electronic device according to any one of claims 1-2, wherein the one or more processors are configured to determine one or more male biomarkers based on the male data, and wherein to provide an output comprises to provide a male output associated with the one or more male biomarkers.

4. Electronic device according to any one of claims 1-3, wherein the male data comprises male blood data of the male, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the male blood data.

5. Electronic device according to any one of claims 1-4, wherein the male data comprises male biosample data of the male, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the male biosample data.

6. Electronic device according to any one of claims 1-5, wherein the female data comprises female blood data of the female, and wherein to determine the one or more riskscores comprises to determine the one or more risk scores based on the female blood data.

7. Electronic device according to any one of claims 1-6, wherein the female data comprises female biosample data of the female, and wherein to determine the one or more risk scores comprises to determine the one or more risk scores based on the female biosample data.

8. Electronic device according to any one of claims 1-7, wherein to determine the one or more risk scores comprises to apply a machine learning model.

9. Electronic device according to any one of claims 1-8, wherein the first risk score comprises a first primary risk score indicative of a possibility of a euploid loss.

10. Electronic device according to any one of claims 1-9, wherein the first risk score comprises a first secondary risk score indicative of a possibility of an aneuploid loss.

11. A computer-implemented method for pregnancy loss classification, the method comprising: obtaining male data associated with a male; obtaining female data associated with a female; obtaining pregnancy loss data associated with a fetus and / or fetal tissue; determining one or more risk scores based on one or more of the male data, the female data, and the pregnancy loss data, the one or more risk scores associated with pregnancy loss and including a first risk score indicative of a fetal ploidy status; and providing an output including a first output associated with the one or more risk scores.

12. A computer-implemented method for training a machine learning model to process as inputs male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue, and provide as output one or more risk scores associated with pregnancy loss, the method comprising:obtaining, using at least one processor, male data associated with a male; female data associated with a female; and pregnancy loss data associated with a fetus and / or fetal tissue; performing, using the at least one processor, a training comprising: generating, using the at least one processor and a machine-learning model, risk data based on the male data, the female data, and the pregnancy loss data; obtaining, using the at least one processor, training data; determining, using the at least one processor and one or more loss functions, one or more loss parameters based on the male data, the female data, the pregnancy loss data, and the training data; and training, using the at least one processor, the machine learning model based on the one or more loss parameters.

Images