System and method for predicting a probability of a pregnant subject with at least one pregnancy complication developing clinical deterioration during a specific time frame

Abstract

The presently disclosed subject matter discloses a system and method for predicting a probability of a pregnant subject with Hypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP -related adverse events during a specific time frame. The system and method utilize at least one of: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of the respective feature, over time, and (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with the pregnant subject or the pregnant subject's fetus, to predict the probability.

Description

[0001] SYSTEM AND METHOD FOR PREDICTING A PROBABILITY OF A PREGNANT SUBJECT WITH AT LEAST ONE PREGNANCY COMPLICATION DEVELOPING CLINICAL DETERIORATION DURING A SPECIFIC TIME FRAME

[0002] TECHNICAL FIELD

[0003] The present invention relates to the field of pregnancy complications, and more particularly, to systems and methods for predicting a probability of a pregnant subject with at least one pregnancy complication developing clinical deterioration within a defined time frame.

[0004] BACKGROUND

[0005] Hypertensive disorders of pregnancy (HDP) refer to a spectrum of conditions characterized by elevated blood pressure during pregnancy, which may include chronic hypertension, gestational hypertension, preeclampsia, and eclampsia. HDP may arise during pregnancy, labor, or the postpartum period, and severity may range from mild to rapidly progressive disease. Without timely recognition and intervention, HDP may be associated with serious maternal and fetal / neonatal complications, including seizures, stroke, placental abruption, medically indicated preterm delivery, maternal morbidity, and perinatal morbidity and mortality.

[0006] HDP complicates substantial proportion of pregnancies (5-8%) and remains a leading cause of maternal and neonatal morbidity and mortality. In some regions, the incidence of HDP has increased in association with a rising prevalence of comorbidities and risk factors, including obesity and chronic hypertension.

[0007] Among pregnant subjects diagnosed with HDP, a subset will progress to preeclampsia with severe features and / or experience adverse events. In clinical practice, expeditious delivery may reduce maternal risk in the presence of severe disease, but delivery prior to term may increase neonatal morbidity associated with prematurity. Accordingly, clinical guidelines may recommend expectant management for certain subjects with preterm HDP who do not present with severe features, with delivery planned at or near a specified gestational age absent disease progression.Since many subjects managed expectantly do not progress to severe disease, the intensity and setting of surveillance may be optimized by tailoring care to individualized risk of progression and / or adverse events. In practice, surveillance options may include outpatient monitoring (e.g., scheduled visits to high-risk obstetric clinics), inpatient admission for intensive monitoring, and intermediate or “hybrid” models that combine remote monitoring with periodic in-person assessment. However, selecting an appropriate surveillance level for a given subject remains challenging due to limited ability to accurately predict near-term deterioration.

[0008] Some approaches use angiogenic biomarkers to stratify risk of progression in pregnant subjects with HDP. For example, the ratio of soluble Fms-like tyrosine kinase- 1 (sFlt-1) to placental growth factor (Pl GF) has been investigated as a predictor of progression to severe disease. Commercial assays have been authorized in certain jurisdictions to aid in assessing risk of progression in selected patient populations. While such biomarker-based approaches can be clinically useful, they may have limitations that reduce generalizability and / or predictive performance in real-world settings. By way of example, performance may vary across populations (e.g., multiple gestations), in subjects receiving particular therapies (e.g., anticoagulation), and across age groups, and reported predictive metrics may be insufficient to minimize falsepositive or false-negative classification in some clinical contexts.

[0009] In addition, certain existing approaches are predominantly “static,” relying on measurements obtained at a single time point or limited time points. Such approaches may not incorporate longitudinal information, including trends in clinical variables and biomarkers over time, nor data captured at higher frequency (e.g., daily, continuous, or near-continuous monitoring). Incorporating longitudinal and multi-source data may enable more accurate and clinically actionable risk estimates within a defined time frame.

[0010] Considering the above, there is a need for a new system and method for predicting a probability of a pregnant subject with at least one pregnancy complication developing clinical deterioration during a specific time frame.

[0011] GENERAL DESCRIPTION

[0012] In accordance with a first aspect of the presently disclosed subject matter, there is provided a system for predicting a probability of a pregnant subject withHypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame, the system comprising a processing circuitry configured to: obtain (a) one or more features associated with the pregnant subject, wherein the features include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of the respective feature, over time, or (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with the pregnant subject or the pregnant subject's fetus, and (b) a machine learning model or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame; and, provide, utilizing the obtained one or more features and the obtained machine learning model, the probability of the pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during the specific time frame.

[0013] In some cases, the machine learning model is trained based on labeled training data containing a plurality of records, wherein each given record contains (i) a plurality of values, each of which is associated with a respective feature of the one or more features, and (ii) a target value configured to indicate whether a respective pregnant subject with HDP, associated with the given record, has develop one or more preeclampsia-related severity features or HDP-related adverse events during the given time frame.

[0014] In some cases, the training data is a dynamic training data capable of being updated continuously or periodically.

[0015] In some cases, instead of obtaining of the features, the system is configured to receive data associated with the pregnant subject and extract the features from the data.

[0016] In some cases, the data includes one or more of: clinical data, laboratory data, and imaging data.

[0017] In some cases, the HDP is one of: gestational hypertension, preeclampsia, preeclampsia with severe features, eclampsia, HELLP syndrome, chronic (preexisting) hypertension, and chronic hypertension with superimposed preeclampsia.

[0018] In some cases, the preeclampsia-related severity features include at least one of: severe blood pressure measurements, symptoms of central nervous system dysfunction,hepatic abnormality, thrombocytopenia, kidney function impairment, and pulmonary edema.

[0019] In some cases, the HDP-related adverse events include at least one of: abruption, eclamptic seizure, Disseminated Intravascular Coagulation (DIC), cerebral hemorrhage / stroke, indicated delivery < 37+0, Birth weight percentile < 10, and fetal / neonatal death.

[0020] In some cases, additionally to providing the probability, the system is configured to identify contributory factors of the features that contributed to the probability.

[0021] In some cases, the one or more features are obtained periodically.

[0022] In some cases, the specific time frame is one of: a day ahead, two days ahead, three days ahead, a week ahead, and two weeks ahead.

[0023] In some cases, the specific timelapse- or imaging-related data includes at least one of: images of fetal ultrasound, fetal heart rate tracing, fetal ECG, continuous maternal vital signs assessment by wearables, or a combination thereof.

[0024] In some cases, the ensemble of models includes at least two of: Random Forest, Logistic Regression, LASSO, and XGBoost.

[0025] In accordance with a second aspect of the presently disclosed subject matter, there is provided a method for predicting a probability of a pregnant subject with Hypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame comprising: obtaining (a) one or more features associated with the pregnant subject, wherein the features include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of the respective feature, over time, or (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with the pregnant subject or the pregnant subject's fetus, and (b) a machine learning model or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame; and, providing, utilizing the obtained one or more features and the obtained machine learning model, the probability of the pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during the specific time frame.In some cases, the machine learning model is trained based on labeled training data containing a plurality of records, wherein each given record contains (i) a plurality of values, each of which is associated with a respective feature of the one or more features, and (ii) a target value configured to indicate whether a respective pregnant subject with HDP, associated with the given record, has develop one or more preeclampsia-related severity features or HDP-related adverse events during the given time frame.

[0026] In some cases, the training data is a dynamic training data capable of being updated continuously or periodically.

[0027] In some cases, instead of obtaining of the features, the method is capable of receiving data associated with the pregnant subject and extracting the features from the data.

[0028] In some cases, the data include one or more of: clinical data, laboratory data, and imaging data.

[0029] In some cases, the HDP is one of: gestational hypertension, preeclampsia, preeclampsia with severe features, eclampsia, HELLP syndrome, chronic (preexisting) hypertension, and chronic hypertension with superimposed preeclampsia.

[0030] In some cases, the preeclampsia-related severity features include at least one of: symptoms of central nervous system dysfunction, hepatic abnormality, thrombocytopenia, kidney function impairment, and pulmonary edema.

[0031] In some cases, the preeclampsia-related adverse events include at least one of: abruption, eclamptic seizure, Disseminated Intravascular Coagulation (DIC), and cerebral hemorrhage / stroke.

[0032] In some cases, additionally to providing the probability, the method is configured to identify contributory factors of the features that contributed to the probability.

[0033] In some cases, the specific time frame is one of: a day ahead, two days ahead, three days ahead, a week ahead, and two weeks ahead.

[0034] In some cases, the one or more features are obtained periodically.

[0035] In some cases, the specific timelapse- or imaging-related data includes at least one of: images of fetal ultrasound, fetal heart rate tracing, fetal ECG, continuous maternal vital signs assessment by wearables, or a combination thereof.In some cases, the ensemble of models includes at least two of: Random Forest, Logistic Regression, LASSO, and XGBoost.

[0036] In accordance with a third aspect of the presently disclosed subject matter, there is provided a non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processor to perform a method for estimating a probability of a pregnant subject with Hypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame, the method for estimating a probability of a pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame comprising: obtaining (a) one or more features associated with the pregnant subject, wherein the features include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of the respective feature, over time, or (ii) at least one highdimensional feature configured to represent specific timelapse- or imaging-related data associated with the pregnant subject or the pregnant subject's fetus, and (b) a machine learning model or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame; and, providing, utilizing the obtained one or more features and the obtained machine learning model, the probability of the pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during the specific time frame.

[0037] BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to understand the presently disclosed subject matter and to see how it may be carried out in practice, the subject matter will now be described, by way of nonlimiting examples only, with reference to the accompanying drawings, in which:

[0039] Fig- 1 illustrates an exemplary table defining the severity features of preeclampsia according to the criteria established by the ACOG, in accordance with the presently disclosed subject matter;

[0040] Fig- 2 illustrates an exemplary table defining preeclampsia adverse events, in accordance with the presently disclosed subject matter;Fig- 3 illustrates an exemplary table of the average Area Under the ROC Curve (AUC) of the algorithm in the repeated cross-validation, in accordance with the presently disclosed subject matter;

[0041] Fig. 4 illustrates the importance of each predictor according to the Random Forest algorithm when fitted on the entire data, in accordance with the presently disclosed subject matter;

[0042] Fig- 5 illustrates an exemplary table of ongoing data collected, in accordance with the presently disclosed subject matter;

[0043] Fig- 6 illustrates an exemplary table of the average AUC of the algorithms predicting developed sPE within the next 3 days, updating every three days, compared to sFlt-1 / Plgf prediction two weeks ahead (see Fig 3), in accordance with the presently disclosed subject matter;

[0044] Fig. 7 is a block diagram schematically illustrating one example of a system for predicting a probability of a pregnant subject with HDP developing preeclampsia-related severity features or HDP -related adverse events during a specific time frame, in accordance with the presently disclosed subject matter; and,

[0045] Fig. 8 is an exemplary flowchart illustrating an example of a sequence of operations carried out by a system for predicting a probability of a pregnant subject with HDP developing preeclampsia-related severity features or HDP-related adverse events during a specific time frame, in accordance with the presently disclosed subject matter.

[0046] DETAILED DESCRIPTION

[0047] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the presently disclosed subject matter. However, it will be understood by those skilled in the art that the presently disclosed subject matter may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the presently disclosed subject matter.

[0048] In the drawings and descriptions set forth, identical reference numerals indicate those components that are common to different embodiments or configurations.

[0049] Unless specifically stated otherwise, as apparent from the following discussions, it is appreciated that throughout the specification discussions utilizing terms such as “obtaining”, “providing”, “training”, “extracting”, “identifying”, or the like, includeaction and / or processes of a computer that manipulate and / or transform data into other data, said data represented as physical quantities, e.g., such as electronic quantities, and / or said data representing the physical objects. The terms “computer”, “processor”, “processing resource”, “processing circuitry”, and “controller” should be expansively construed to cover any kind of electronic device with data processing capabilities, including, by way of non-limiting example, a personal desktop / laptop computer, a server, a computing system, a communication device, a smartphone, a tablet computer, a smart television, a processor (e.g. digital signal processor (DSP), a microcontroller, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), etc.), a group of multiple physical machines sharing performance of various tasks, virtual servers co-residing on a single physical machine, any other electronic computing device, and / or any combination thereof.

[0050] The operations in accordance with the teachings herein may be performed by a computer specially constructed for the desired purposes or by a general-purpose computer specially configured for the desired purpose by a computer program stored in a non-transitory computer readable storage medium. The term "non-transitory" is used herein to exclude transitory, propagating signals, but to otherwise include any volatile or non-volatile computer memory technology suitable to the application.

[0051] As used herein, the phrase "for example," "such as", "for instance" and variants thereof describe non-limiting embodiments of the presently disclosed subject matter. Reference in the specification to "one case", "some cases", "other cases" or variants thereof means that a particular feature, structure or characteristic described in connection with the embodiment(s) is included in at least one embodiment of the presently disclosed subject matter. Thus, the appearance of the phrase "one case", "some cases", "other cases" or variants thereof does not necessarily refer to the same embodiment(s).

[0052] It is appreciated that, unless specifically stated otherwise, certain features of the presently disclosed subject matter, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the presently disclosed subject matter, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination.In embodiments of the presently disclosed subject matter, fewer, more and / or different stages than those shown in Fig. 8 may be executed. In embodiments of the presently disclosed subject matter one or more stages illustrated in Fig. 8 may be executed in a different order and / or one or more groups of stages may be executed simultaneously. Each module in Fig. 7 can be made up of any combination of software, hardware and / or firmware that performs the functions as defined and explained herein. The modules in Fig. 7 may be centralized in one location or dispersed over more than one location. In other embodiments of the presently disclosed subject matter, the system may comprise fewer, more, and / or different modules than those shown in Fig. 7.

[0053] Any reference in the specification to a method should be applied mutatis mutandis to a system capable of executing the method and should be applied mutatis mutandis to a non-transitory computer readable medium that stores instructions that once executed by a computer result in the execution of the method.

[0054] Any reference in the specification to a system should be applied mutatis mutandis to a method that may be executed by the system and should be applied mutatis mutandis to a non-transitory computer readable medium that stores instructions that may be executed by the system.

[0055] Any reference in the specification to a non-transitory computer readable medium should be applied mutatis mutandis to a system capable of executing the instructions stored in the non-transitory computer readable medium and should be applied mutatis mutandis to method that may be executed by a computer that reads the instructions stored in the non-transitory computer readable medium.

[0056] By way of introduction, the presently disclosed subject matter relates to a maternal-fetal monitoring and prediction system configured to provide continuous, individualized, and dynamic prediction of clinical deterioration in pregnant subjects affected by pregnancy complications. Such pregnancy complications include, but are not limited to, hypertensive disorders of pregnancy (HDP), preterm pre-labor rupture of membranes (PPROM), preterm birth, intrahepatic cholestasis of pregnancy, intrauterine growth restriction (IUGR), monochorionic diamniotic (MCDA) twin pregnancies, placenta accreta spectrum, placenta previa, vasa previa, and complications arising from pre-existing maternal disease.

[0057] In some embodiments, the system may comprise a subject-specific model of a maternal-fetal unit being continuously updated from initial recognition of a pregnancycomplication and throughout the remainder of the pregnancy. In contrast to conventional risk stratification tools that produce static predictions at predefined time points, the system provides ongoing probabilistic risk estimates that evolve over time as new information is received. The risk estimates may incorporate a current clinical state as well as temporal trends, trajectories, and derived features computed from longitudinal, multimodal data sources, including one or more of: clinical complaints and symptoms; maternal vital signs; laboratory results; fetal heart rate monitoring (including non-stress testing); and ultrasound-based fetal biometric, biophysical, and Doppler measurements.

[0058] For a given pregnancy complication, the system may be configured to predict an anticipated course of disease and / or deterioration, including both: (i) complicationspecific outcomes that are commonly recognized for the pregnancy complication; and (ii) emerging, secondary, or evolving risks that may not be directly attributable to the initial diagnosis but may be inferred from evolving maternal and / or fetal physiologic and pathophysiologic patterns reflected in the longitudinal data.

[0059] By way of non-limiting examples, the predicted adverse outcomes may include one or more of:

[0060] (a) for HDP: eclampsia, hemolysis-elevated liver enzymes-low platelets (HELLP) syndrome, development of severe features of preeclampsia, and placental abruption;

[0061] (b) for PPROM: preterm birth and chorioamnionitis;

[0062] (c) for IUGR: intrauterine fetal death and perinatal asphyxia;

[0063] (d) for intrahepatic cholestasis of pregnancy: intrauterine fetal death, perinatal asphyxia, and preterm birth; and

[0064] (e) for MCDA twin pregnancy: intrauterine fetal death, perinatal asphyxia, preterm birth, and fetal brain injury.

[0065] It is to be of note that the above examples are provided for explanatory purposes and are not intended in any way to limit the scope of the present disclosure. Other adverse outcomes may be predicted depending on the pregnancy complication, population, and clinical setting.

[0066] In some embodiments, the system may implement multimodal longitudinal learning using one or more trained predictive models that ingest longitudinal clinical data, including, without limitation: maternal vital signs; laboratory results (includinghematologic and biochemical markers, angiogenic biomarkers, and urine testing); fetal surveillance data (including non-stress tests and related fetal monitoring outputs); and imaging data (including ultrasound and Doppler studies). Rather than treating measurements as isolated data points, the system may analyze temporal trajectories, enabling identification of evolving patterns that precede maternal and / or fetal deterioration.

[0067] Key distinguishing features of the presently disclosed subject matter may include one or more of: dynamic, time-continuous prediction rather than single timepoint risk assessment; adaptive learning from within-subject changes across a pregnancy timeline; integration of maternal and fetal data into a unified computational model; and prediction of imminent clinical deterioration and / or adverse outcomes, rather than retrospective diagnosis or static classification. By modeling maternal and fetal physiology as a continuously evolving system, the disclosed approach supports earlier identification of impending complications, facilitates proactive and personalized clinical decision-making, and enables timely intervention to improve maternal and neonatal outcomes.

[0068] In some embodiments, the presently disclosed subject matter may further include an operational unit-level computational model configured to aggregate outputs generated by a plurality of individual subject-specific models associated with respective pregnant subjects monitored by a maternal-fetal medicine unit. The unit-level computational model may enable real-time and retrospective assessment of unit-level risk and resource utilization, thereby supporting clinical and administrative efficiency. For example, the unit-level computational model may support workforce and staffing planning, anticipation of urgent admissions, labor and delivery capacity management, and evaluation of clinical, operational, and cost-effectiveness impacts of protocol changes over time. This dual-layer architecture-combining individualized patient-level modeling with unit-level operational modeling, extends the presently disclosed subject matter beyond patient-specific prediction to healthcare system optimization in maternal-fetal medicine.

[0069] It is to be of note throughout the description that references made to “Hypertensive Disorders of Pregnancy (HDP)” and / or to “preeclampsia-related severity features or HDP -related adverse events” are provided solely as illustrative examples to facilitate better understanding of the presently disclosed subject matter. Such referencesare not intended to, and shall not be construed to, limit the scope of the disclosed subject matter, which may be applied to other pregnancy-related conditions, complications, severity features, adverse events, or clinical deterioration scenarios within a given time frame.

[0070] As a first aspect of the presently disclosed subject matter, the presently disclosed subject matter leverages advanced machine learning models to integrate clinical, lab, and imaging data to enhance the accuracy of the prediction for both sPE (several examples of which are listed in Fig. 1) and related adverse events (several examples of which are listed in Fig. 2) within a time frame of, for example, two weeks ahead.

[0071] It is to be of note that the specific sPE and related adverse events mentioned in both Fig. 1 and Fig. 2 serve as mere examples not intended in any way to limit the scope of the presently disclosed subject matter, and that other types of both sPE and related adverse events may also be applicable, mutatis mutandis.

[0072] It is to be further of note that the time frame of two weeks mentioned above serves as a mere example not intended in any way to limit the scope of the presently disclosed subject matter, and that other time frames (e.g., a day ahead, two days ahead, three days ahead, a week ahead, etc.) may also be applicable, mutatis mutandis.

[0073] In accordance with the above, by way of a non-limiting example, presented merely for the purpose of better understanding the presently disclosed subject matter and not intended in any way to limit its scope, data were collected from pregnant subjects admitted to an antepartum high-risk unit prior to 35 weeks of gestation for HDP, with angiogenic biomarkers measured including sFlt-1 and P1GF. Enrollment time was defined based on the time of serum sFIt-l / PIGF testing. After enrollment, an eligible subject could be re-enrolled after completion of the prediction horizon if the subject had not yet delivered. In this example, there were 88 total entries, and 31 entries (35%) were associated with progression to sPE within the applicable horizon.

[0074] Within the above example, multiple predictive approaches were evaluated using repeated cross-validation (e.g., repeated 5-fold cross-validation). Non-limiting algorithm examples include penalized regression (e.g., LASSO), tree-based models (e.g., random forest), boosted tree models, and logistic regression models (including stepwise variants). It is to be of note that these algorithms serve as mere examples not intended in any way to limit the scope of the presently disclosed subject matter, and that other algorithms known in the art may also be applicable, mutatis mutandis.In some embodiments, the model inputs include one or more of: maternal demographics and medical history; pregnancy characteristics (including plurality); medication exposures (including antiplatelet therapy and anticoagulation); vital signs (including systolic and diastolic blood pressure); laboratory values (including complete blood count parameters, liver enzymes, renal function markers, uric acid, and proteinuria measures such as 24-hour urine protein and / or urine protein-to-creatinine ratio); fetal biometry; fetal Doppler indices; angiogenic biomarkers (including sFlt-1, P1GF, and ratios or transforms thereof); and combinations of the foregoing. It is to be of note that the above serve as mere examples not intended in any way to limit the scope of the presently disclosed subject matter, and that other features may also be applicable, mutatis mutandis.

[0075] In some embodiments, predictive performance is evaluated using one or more discrimination and / or calibration metrics, such as area under the receiver operating characteristic curve (AUC), and may be compared against one or more baselines, including biomarker-only baselines (e.g., sFIt-l / PIGF ratio, see Fig. 3). In some embodiments, model interpretability outputs such as feature importance, attribution, or sensitivity analyses are generated (e.g., as illustrated in Fig. 4 for a tree-based model).

[0076] As a second aspect of the presently disclosed subject matter, the current invention provides dynamic prediction, in which the risk estimate is updated as new measurements become available. For example, the probability of developing sPE and / or an adverse event may be re-estimated upon receipt of each new data point or each new batch of data collected over time (e.g., as described in Fig. 5). It is to be of note that the any data mentioned in Fig. 5 serves as a mere example not intended in any way to limit the scope of the presently disclosed subject matter, and that other types of data may also be applicable, mutatis mutandis.

[0077] In some embodiments, dynamic models, such as LASSO, Random Forrest, boosting trees, and stepwise logistic regression, may be configured to output risk for a near-term horizon (e.g., the next 24 hours, 48 hours, 72 hours, or another interval), and the updating may be performed at a selected cadence (e.g., daily, every 3 days, continuously, or when new clinically relevant measurements are recorded). Fig. 6 illustrates the AUC of the algorithms predicting developed sPE within the next 3 days, updating every three days, compared to sFlt- 1 / Plgf prediction two weeks ahead.It is to be of note that the above mentioned algorithms serve as mere examples not intended in any way to limit the scope of the presently disclosed subject matter, and that other algorithms known in the art may also be applicable, mutatis mutandis.

[0078] As a third aspect of the presently disclosed subject matter, the current invention incorporates advanced data types to improve prediction, including one or more of: ultrasound images, fetal heart rate tracings, fetal electrocardiography (ECG), maternal physiologic monitoring from wearable sensors, and other high-frequency or continuous data streams. In some embodiments, computer vision techniques are used to extract features from ultrasound images; signal-processing and machine learning techniques (including deep learning) are used to extract features from fetal heart rate tracings and / or fetal ECG; and time-series modeling techniques are used to incorporate continuous or near-continuous maternal vital signs.

[0079] As a fourth aspect of the presently disclosed subject matter, the current invention uses changes over time (“delta” features), trends, rates of change, and / or variability measures derived from longitudinal measurements to refine risk estimates.

[0080] Non-limiting examples include deltas and trends in: fetal biometric parameters including biparietal diameter, head circumference, abdominal circumference, femur length and the estimated fetal weight calculated from them, fetal doppler flows, maternal systolic and diastolic blood pressure, complete blood count features including platelet level, LDH, ALT, AST, creatinine level, uric acid, urine protein collected 24 hours or urine protein to creatinine ratio (UPCR) as well sFlt-1, Pl GF and the sFlt-1 / Plgf ratio. It is to be of note that the features mentioned above serve as mere examples not intended in any way to limit the scope of the presently disclosed subject matter, and that other features may also be applicable, mutatis mutandis.

[0081] As a fifth aspect of the presently disclosed subject matter, the current invention uses an ensemble of predictive models to generate the risk estimate. For example, an ensemble may combine outputs of one or more of: logistic regression, penalized regression (e.g., LASSO), tree-based models (e.g., random forest), and gradient boosting models (e.g., XGBoost). In some embodiments, ensemble weighting is determined based on the data modality, the prediction horizon, and / or performance on a validation set, and may be configured to improve robustness and reduce overfitting.It is to be of note that the models mentioned above serve as mere examples not intended in any way to limit the scope of the presently disclosed subject matter, and that other models may also be applicable, mutatis mutandis.

[0082] In some embodiments, integration of imaging and tabular data is performed via multi-modal fusion, including feature-level fusion, decision-level fusion, and / or hybrid fusion approaches. In some embodiments, time-series and “time-lapse” data are incorporated using models configured for high-frequency measurements (e.g., blood pressure and fetal heart rate surveillance), using both absolute values and temporal derivatives. In some embodiments, model selection and / or model complexity is adapted based on dataset characteristics, for example using landmarking or related meta-learning approaches. In some embodiments, ensembles include weighted averaging, stacking (including a meta-model trained on model outputs), blending, voting (hard and / or soft), and / or cascading architectures. In some embodiments, feature engineering is shared across models, such as using feature importance from a tree-based model to select or weight features for a regression-based model.

[0083] Attention is now drawn to Fig. 7, depicting a block diagram schematically illustrating one example of the system for predicting the probability of a pregnant subject with HDP developing preeclampsia-related severity features or HDP-related adverse events during a specific time frame, in accordance with the presently disclosed subject matter.

[0084] In accordance with the presently disclosed subject matter, the system for predicting the probability of a pregnant subject with HDP developing preeclampsia-related severity features or HDP-related adverse events during a specific time frame 100 (also interchangeably referred to herein as “system 100”) may comprise a network interface 106. The network interface 106 (e.g., a network card, a Wi-Fi client, 3G / 4G client, or any other component), enables system 100 to communicate over a network with external systems and handles inbound and outbound communications from such systems. For example, system 100 may receive, through network interface 106, one or more features associated with a pregnant subject with HDP.

[0085] System 100 may further comprise or be otherwise associated with a data repository 104 (e.g., a database, a storage system, a memory including Read Only Memory - ROM, Random Access Memory - RAM, or any other type of memory, etc.)configured to store data. Some examples of data that may be stored in the data repository 104 include:

[0086] • One or more features associated with one or more pregnant women with HDP;

[0087] • One or more delta features;

[0088] • One or more high-dimensional features;

[0089] • One or more machine learning models or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame;

[0090] • One or more labeled training data containing a plurality of records, each associated with a given pregnant subject with HDP;

[0091] • One or more probabilities of one or more women with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during said specific time frame;

[0092] • One or more contributory factors of features associated with one or more pregnant women with HDP; etc.

[0093] Data repository 104 may be further configured to enable retrieval and / or update and / or deletion of the stored data. It is to be noted that in some cases, data repository 104 may be distributed, while the system 100 has access to the information stored thereon, e.g., via a wired or wireless network to which system 100 is able to connect (utilizing its network interface 106).

[0094] System 100 further comprises processing circuitry 102. Processing circuitry 102 may be one or more processing units (e.g., central processing units), microprocessors, microcontrollers (e.g., microcontroller units (MCUs)) or any other computing devices or modules, including multiple and / or parallel and / or distributed processing units, which are adapted to independently or cooperatively process data for controlling relevant system 100 resources and for enabling operations related to system’s 100 resources.

[0095] The processing circuitry 102 comprises a probability prediction module 108, configured to perform a probability prediction process, as further detailed herein, inter alia with reference to Fig. 8.Turning to Fig. 8 there is shown a flowchart illustrating one example of operations carried out by the system for predicting the probability of a pregnant subject with pregnant subject with HDP developing preeclampsia-related severity features or pregnant subject with HDP-related adverse events during a specific time frame 100, in accordance with the presently disclosed subject matter.

[0096] Accordingly, the system for predicting or estimating the probability of a pregnant subject with HDP developing preeclampsia-related severity features or pregnant subject with HDP-related adverse events during a specific time frame 100 (also interchangeably referred to hereafter as “system 100”) may be configured to perform a probability prediction process 200, e.g., using probability prediction module 108.

[0097] For this purpose, system 100 obtains (a) one or more features associated with a pregnant subject with HDP, and (b) a machine learning model capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame (block 202).

[0098] In some cases, the one or more features may include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of said respective feature, over time, or (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with said pregnant subject or said pregnant subject's fetus.

[0099] In some cases, instead of obtaining of said features, system 100 may be configured to receive data associated with said pregnant subject and extract said features from said data, using any appropriate method known in the art for such purpose. In such cases, the data may include one or more of: clinical data, laboratory data, and imaging data.

[0100] In some cases, the machine learning model may be trained based on labeled training data containing a plurality of records. Each record may contain: (i) a plurality of values, each of which is associated with a respective feature of the one or more features, and (ii) a target value configured to indicate whether a respective pregnant subject with HDP, associated with the given record, has develop one or morepreeclampsia-related severity features or HDP-related adverse events during the given time frame.

[0101] In some cases, the training data may be a dynamic training data capable of being updated continuously or periodically.

[0102] In some cases, the HDP may be one of: gestational hypertension, preeclampsia, preeclampsia with severe features, eclampsia, HELLP syndrome, chronic (preexisting) hypertension, chronic hypertension with superimposed preeclampsia, chronic hypertension with superimposed preeclampsia with severe features, etc.

[0103] In some cases, the preeclampsia-related severity features may include at least one of: severe blood pressure measurements, symptoms of central nervous system dysfunction, hepatic abnormality, thrombocytopenia, kidney function impairment, pulmonary edema, etc.

[0104] In some cases, the preeclampsia-related adverse events may include at least one of: abruption, eclamptic seizure, Disseminated Intravascular Coagulation (DIC), cerebral hemorrhage / stroke, indicated delivery < 37+0, Birth weight percentile < 10, fetal / neonatal death, etc.

[0105] In some cases, the one or more features may be obtained periodically (every three hours, every six hours, every three days, etc.).

[0106] Once obtained, system 100 provides, utilizing the obtained one or more features and the obtained machine learning model, the probability of the pregnant subject with preeclampsia to develop one or more preeclampsia-related severity features or preeclampsia-related adverse events during said specific time frame (block 204).

[0107] In some cases, additionally to providing the probability, system 100 may be configured to identify contributory factors of the features that contributed to the probability.

[0108] In some cases, the specific time frame may be one of: a day ahead, two days ahead, three days ahead, two weeks ahead, etc.

[0109] It is to be noted, with reference to Fig- 8, that some of the blocks can be integrated into a consolidated block or can be broken down to a few blocks and / or other blocks may be added. It is to be further noted that some of the blocks are optional. It should be also noted that whilst the flow diagram is described also with reference to the system elements that realizes them, this is by no means binding, and the blocks can be performed by elements other than those described herein.It is to be understood that the presently disclosed subject matter is not limited in its application to the details set forth in the description contained herein or illustrated in the drawings. The presently disclosed subject matter is capable of other embodiments and of being practiced and carried out in various ways. Hence, it is to be understood that the phraseology and terminology employed herein are for the purpose of description and should not be regarded as limiting. As such, those skilled in the art will appreciate that the conception upon which this disclosure is based may readily be utilized as a basis for designing other structures, methods, and systems for carrying out the several purposes of the present presently disclosed subject matter.

[0110] It will also be understood that the system according to the presently disclosed subject matter can be implemented, at least partly, as a suitably programmed computer. Likewise, the presently disclosed subject matter contemplates a computer program being readable by a computer for executing the disclosed method. The presently disclosed subject matter further contemplates a machine-readable memory tangibly embodying a program of instructions executable by the machine for executing the disclosed method.

Claims

CLAIMS:

1. A system for predicting a probability of a pregnant subject with Hypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame, the system comprising a processing circuitry configured to:obtain (a) one or more features associated with said pregnant subject, wherein said features include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of said respective feature, over time, or (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with said pregnant subject or said pregnant subject's fetus, and (b) a machine learning model or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame; and, provide, utilizing the obtained one or more features and the obtained machine learning model, the probability of said pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during said specific time frame.

2. The system of claim 1, wherein said machine learning model is trained based on labeled training data containing a plurality of records, wherein each given record contains (i) a plurality of values, each of which is associated with a respective feature of said one or more features, and (ii) a target value configured to indicate whether a respective pregnant subject with HDP, associated with said given record, has develop one or more preeclampsia-related severity features or HDP-related adverse events during said given time frame.

3. The system of claim 2, wherein said training data is a dynamic training data capable of being updated continuously or periodically.

4. The system of claim 1, wherein, instead of obtaining of said features, said system is configured to receive data associated with said pregnant subject and extract said features from said data.

5. The system of claim 3, wherein said data include one or more of: clinical data, laboratory data, and imaging data.

6. The system of claim 1, wherein said HDP is one of: gestational hypertension, preeclampsia, preeclampsia with severe features, eclampsia, HELLP syndrome, chronic (preexisting) hypertension, and chronic hypertension with superimposed preeclampsia.

7. The system of claim 1, wherein said preeclampsia-related severity features include at least one of: severe blood pressure measurements, symptoms of central nervous system dysfunction, hepatic abnormality, thrombocytopenia, kidney function impairment, and pulmonary edema.

8. The system of claim 1, wherein said HDP -related adverse events include at least one of: abruption, eclamptic seizure, Disseminated Intravascular Coagulation (DIC), cerebral hemorrhage / stroke, indicated delivery < 37+0, Birth weight percentile < 10, and fetal / neonatal death.

9. The system of claim 1, wherein, additionally to providing said probability, said system is configured to identify contributory factors of said features that contributed to said probability.

10. The system of claim 1, wherein said one or more features are obtained periodically.

11. The system of claim 1, wherein said specific time frame is one of: a day ahead, two days ahead, three days ahead, a week ahead, and two weeks ahead.

12. The system of claim 1, wherein said specific timelapse- or imaging-related data includes at least one of: images of fetal ultrasound, fetal heart rate tracing, fetalECG, continuous maternal vital signs assessment by wearables, or a combination thereof.

13. The system of claim 1, wherein said ensemble of models include at least two of:Random Forest, Logistic Regression, LASSO, and XGBoost.

14. A method for predicting a probability of a pregnant subject with Hypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame comprising:obtaining (a) one or more features associated with said pregnant subject, wherein said features include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of said respective feature, over time, or (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with said pregnant subject or said pregnant subject's fetus, and (b) a machine learning model or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame; and, providing, utilizing the obtained one or more features and the obtained machine learning model, the probability of said pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during said specific time frame.

15. The method of claim 14, wherein said machine learning model is trained based on labeled training data containing a plurality of records, wherein each given record contains (i) a plurality of values, each of which is associated with a respective feature of said one or more features, and (ii) a target value configured to indicate whether a respective pregnant subject with HDP, associated with said given record, has develop one or more preeclampsia-related severity features or HDP-related adverse events during said given time frame.

16. The method of claim 15, wherein said training data is a dynamic training data capable of being updated continuously or periodically.

17. The method of claim 14, wherein, instead of obtaining of said features, said method is capable of receiving data associated with said pregnant subject and extracting said features from said data.

18. The method of claim 17, wherein said data include one or more of: clinical data, laboratory data, and imaging data.

19. The method of claim 14, wherein said HDP is one of: gestational hypertension, preeclampsia, preeclampsia with severe features, eclampsia, HELLP syndrome, chronic (preexisting) hypertension, and chronic hypertension with superimposed preeclampsia.

20. The method of claim 14, wherein said preeclampsia-related severity features include at least one of: symptoms of central nervous system dysfunction, hepatic abnormality, thrombocytopenia, kidney function impairment, and pulmonary edema.

21. The method of claim 14, wherein said preeclampsia-related adverse events include at least one of: abruption, eclamptic seizure, Disseminated Intravascular Coagulation (DIC), and cerebral hemorrhage / stroke.

22. The method of claim 14, wherein, additionally to providing said probability, said method is configured to identify contributory factors of said features that contributed to said probability.

23. The method of claim 14, wherein said specific time frame is one of: a day ahead, two days ahead, three days ahead, a week ahead, and two weeks ahead.

24. The method of claim 14, wherein said one or more features are obtained periodically.

25. The system of claim 14, wherein said specific timelapse- or imaging-related data includes at least one of images of fetal ultrasound, fetal heart rate tracing, fetal ECG, continuous maternal vital signs assessment by wearables, or a combination thereof.

26. The system of claim 14, wherein said ensemble of models include at least two of:Random Forest, Logistic Regression, LASSO, and XGBoost.

27. A non-transitory computer readable storage medium having computer readable program code embodied therewith, the computer readable program code, executable by at least one processor to perform a method for predicting a probability of a pregnant subject with Hypertensive Disorders of Pregnancy (HDP) to develop one or more preeclampsia-related severity features or HDP -related adverse events during a specific time frame, the method for predicting a probability of a pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a specific time frame comprising:obtaining (a) one or more features associated with said pregnant subject, wherein said features include: (i) at least one delta feature configured to indicate a change in a respective feature between different measurements of said respective feature, over time, or (ii) at least one high-dimensional feature configured to represent specific timelapse- or imaging-related data associated with said pregnant subject or said pregnant subject's fetus, and (b) a machine learning model or an ensemble of models capable of receiving one or more features associated with a given pregnant subject with HDP and providing the probability of the given pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during a given time frame; and, providing, utilizing the obtained one or more features and the obtained machine learning model, the probability of said pregnant subject with HDP to develop one or more preeclampsia-related severity features or HDP-related adverse events during said specific time frame.

Images