Method and system for fetal monitoring and treatment

Abstract

A method of treating a mother intrapartum. The method includes obtaining fetal heart rate (FHR) and uterine activity data, including contraction data, from a pregnant mother intrapartum. Contraction data for a plurality of time periods is determined. The FHR data and contraction data is analyzed, within each of the plurality of time periods, to determine occurrence of a marker selected from the markers consisting of Fetal Evaluation, Fetal Intervention, Fetal Operation, and Fetal Morbidity. A recommended intervention is provided based on the occurrence of the marker. The recommendation may be, for example, shown on a display.

Description

Method and System for Fetal Monitoring and TreatmentCross-Reference to Related Applications

[0001] This application claims priority from U.S. Provisional Patent Application 63 / 509,470, filed June 21,2023, which is hereby incorporated herein by reference in its entirety.Technical Field

[0002] The present invention relates to fetal monitoring and treatment of laboring patients, and more particularly to modeling Fetal Heart Rate characteristics during various contraction periods to identify and treat threats to fetal and maternal well-being.Background Art

[0003] In the United States, there are approximately four million documented births per year. Over 85% of those cases utilize electronic fetal monitoring (EFM) at some point during the pregnancy. This is true especially during the labor process, as a means for overseeing fetal and maternal well-being. This monitoring of the fetal heart rate (FHR) pattern and uterine contractions, especially when combined with the information regarding the mother’s well- being and the course of labor contains valuable data about the state of the fetus, threats to oxygenation, to placental, cerebral, and umbilical blood flow, and other threats to the fetus and mother.

[0004] Despite the aforementioned benefits, more than fifty years after its introduction, FHR monitoring continues to provoke debate about its value and especially its role as a stimulus to cesarean sections as well as allegations of medical malpractice. Adverse fetal outcomes are often associated with the failure to timely and / or properly process, acknowledge and / or respond to the FHR pattern information. The misinterpretation and / or ambiguity of the EFM patterns has contributed to the significantly increased use of Cesareandelivery from about 5% in the 1970s to over 30% as of 2016, leading to increased personal and governmental expenditures, costing the country over $1 billion per year per every 5% of additional Cesarean deliveries, in addition to potential adverse effects on the health of the mothers and children. Obstetrical liability costs the country approximately $40 billion per year, of which 70% is accounted for by EFM (mis)interpretation and related birth- related brain injury.

[0005] Review of the accuracy of current FHR evaluation methodologies reveal that approximately 67% (two thirds) of neonates with Hypoxic-Ischemic Encephalopathy (HIE) neurologic injury never reach the definitions of ACOG (American College of Obsterics and Gynecology), nor do forty -percent of HIE babies ever exhibit a Category III tracing (Category System) indicative of neurologic injury during labor. Recent review of a limited number of FHR tracings of fetuses subsequently found to display autistic features has revealed identifiable changes in FHR patterns during labor that mimic the patterns seen with babies with hypoxic-ischemic encephalopathy (HIE).

[0006] Prior to the introduction of current FHR analysis programs (ACOG Category System) in 2003, the incidence of Autism was 1 per 2000 live births. Twenty years after the change from traditional FHR analysis during labor to the current ACOG methodology, the incidence of Autism Syndrome has risen to 1 per 33-50 live births. Current FHR interpretation methodology and practice to date has not addressed the issue of the prevention of autism syndrome resulting from labor despite significant neurological evidence suggesting the link between autism syndrome and labor.

[0007] Therefore, the dramatic rise of both Autism spectrum disorder (ASD) and the apparent failure to lower the risk of Hypoxic Ischemic Encephalopathy (HIE) during the last two decades appears to underscore the significant limitations in the approach taken to the surveillance of the fetus during labor, with its emphasis of fetal hypoxia / acidosis / anoxia and its failure to include non-FHR attributes including progress in labor and the diagnosis of Excessive Uterine Activity (EXU A).

[0008] The simplistic ACOG Category System has resulted in a dramatic reduction in physician, midwife, nurse practitioner education of FHR tracing, emphasizing the need for computer systems using mathematic modeling and visual pattern recognition (Artificial Intelligence) in order to reduce emergency deliveries and irreversible brain damage of the fetus and neonate.

[0009] Fig. 1 is an example of a standard FHR and uterine contraction (UC) time series chart. The chart contains two components: the top is the FHR signal from the fetus, and the bottom component is the UC signal from the mother during labor. The data is typically captured simultaneously. Traditionally, such data is stored electronically as non- vectorized images, for example in a pdf format.

[0010] Currently, there is no mechanism to efficiently extract FHR features from digital charts, aside from semi-automated or manual chart digitization. Some currently available methods rely on extracting graphical data from digital charts such as shown in Fig. 1, through manually selecting points on of the plot(s) and determining the minimum and maximum values of the chart. Typically, the next step involves manually selecting additional points of the plotted data points in the chart, necessary for the automation tools to interpolate the values of interest. Such extraction is prone to human bias, uncertainty, and therefore may result in inaccurate data representation and / or reproduction. The errors might be even more significant when the observed pattern is abnormal, confounded by lost data, and / or the insertion of the maternal heart rate (a common issue associated with electronic fetal heart rate monitoring) thus rendering such tools even more unreliable.

[0011] Generally, separate data points displayed in such charts bear little useful information when observed out of context. Instead, the true value of these charts lies in relative positions of those data points, which can be used as a means to determine data correlations, ascertain baselines, etc. For example, such information could be used to process and / or identify a relevant piece of information (e.g., Is this heartbeat normal? Is the fetus in distress? Is the labor progressing normally?). Therefore, a need exists for a method of comprehensive data analysis, wherein the data is treated based on trends, and not onparticular separate points in the charts. The correlations include not only the condition of the fetus but the prevention of the need for urgent intervention where possible.

[0012] Numerous attempts have been made to apply machine learning techniques for feature recognition, or for automated feature analysis in the field of FHR monitoring, yet none address the need to automatically process information presented in non-digitized charts in general, and to detect systematic features found in the FHR charts in particular. Methods that focus on FHR feature extraction largely rely on manual intervention, thus rendering themselves inefficient and unreliable. Methods that cover machine learning or computer vision usually focus on extracting and reproducing the charts in their entirety rather than on extracting and processing specific features from the charts. When such methods focus on machine learning approaches in application to data extraction, they suffer from severe accuracy and efficiency issues, and therefore cannot be applied to processing substantial sets of data, such as what is recommended for extensive training of an artificial intelligence model in general, and in FHR application in particular. Methods that focus on FHR data processing generally lack a machine learning component altogether.Summary of the Embodiments

[0013] In illustrative embodiments of the invention, a new mathematical and visual recognition system and methodology is provided for directing mathematical calculation and modeling of FHR characteristics during the pre-contraction, contraction, and post-contraction periods during labor to earlier identify threats to fetal reserve and injury. The system and methodology advantageously helps to prevent emergency operative deliveries (vacuum extraction and cesarean delivery), reducing the costs of labor for hospitals and insurance carriers, reducing the excessive need for emergency nursing and physician staffing of labor and delivery to accommodate emergency delivery, and reducing the incidence and costs of permanent fetal neurologic injury (Austism, HIE, and Cerebral Palsy).

[0014] In accordance with an embodiment of the invention, a method of treating a mother intrapartum is provided. The method includes obtaining fetal heart rate (FHR) and uterine activity data, including contraction data, from a pregnant mother intrapartum.Contraction data for each contraction is determined. The contraction data may include a plurality of time periods, for example, at least four time periods including: a first time period defined as predefined time before the contraction begins to the onset of the contraction; a second time period defined as duration of the contraction, a third time period defined as a predefined period after the contraction ends, and a fourth time period defined as any additional time after the end of the third time period till a predetermined time before the next contraction. The FHR data and contraction data is analyzed, within each of the plurality of time periods, to determine occurrence of a marker selected from the markers consisting of Fetal Evaluation, Fetal Intervention, Fetal Operation, and Fetal Morbidity. A recommended intervention is provided based on the occurrence of the marker. The recommendation may be, for example, shown on a display.

[0015] In accordance with related embodiments of the invention, the method may further include performing the recommended intervention. The recommended intervention may be selected from one or more of: no intervention required, improving tracing of the FHR pattern and / or contraction pattern by adjusting or switching the measuring equipment, desisting from pushing and / or desisting from inducing labor, reducing uterine activity by reducing Pitocin and / or administering uterine muscle relaxant, and delivery.

[0016] In accordance with further related embodiments of the invention, the recommended intervention for Fetal Evaluation (FE) may include: examine cervix to determine cervical dilation effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis; perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE); perform artificial rupture of membranes (AROM); reducing or stopping administration of picotin; place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion;if in 2nd stage (pushing), allow pushing only every other contraction, but if nonengaged do not allow push with each contraction; perform abdominal ultrasound to determine present fetal presentation; begin discussion of informed consent for operative vaginal delivery and begin consent process; inform labor staff of the potential need to perform operative intervention to helpassemble the 2nd OR team; and / or evaluate for chori oamnionitis if fetal tachycardia or s / s chori oamnioitis with administration of antibiotics if IAI present or suspected.

[0017] In accordance with still further related embodiments of the invention, the recommended intervention for Fetal Intervention (FI) may include: examine cervix to determine cervical dilation, effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis; perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE); perform artificial rupture of membranes (AROM) if not already done; cease all pitocin induction / augmentation if being done and pushing efforts if in the 2nd stage of labor; place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion; administer 25 cc sc terbutaline if contractions > 6 / 10 minutes or hypertonic uterus; prepare staff for the potential of operative delivery; discuss and signing of operative delivery informed consent; full intrauterine resuscitation (IR) with cessation of all induction agents including pitocin, cervadil and / or cytotec; maternal repositioning, infusion of IV fluids; provide oxygen by mask; mnioinfusion if variable FHR decelerations have not started; move patient to operating room if FI persists more than 10 minutes; and / or prepare and examine patient for possible operative dl by anesthesiologist.

[0018] In accordance with yet further related embodiments of the invention, the recommended intervention for Fetal Operation (FO) includes: prep patient for operative; induce anesthesia if FHR tracing has not improved within 10 minutes; perform operative delivery; fetal head cooling after delivery; continue FHR monitoring following delivery for 1 hour to observe for characteristics of fetal tachycardia and absent variability (Category III tracing); perform c / section; and / or cord gases and placental pathology.

[0019] In accordance with further even further related embodiments of the invention, the recommended intervention for Fetal Morbidity (FM) includes: full neonatal resuscitation with Fetal scalp electrode on the fetus after delivery; emergency delivery; brain cooling, and / or cord gases and placental pathology.

[0020] In accordance with more related embodiments of the invention, analyzing may include determining fetal features including baseline fetal heart rate (FHR), baseline variability, accelerations, decelerations, fetal recovery, suspected fetal acidosis, and / or fetal behavior. Analyzing may include determining uterine contraction parameters including baseline tone, frequency, amplitude, duration, intervals, rest-time, not interpretable, absent, rhythm, pushing, and / or uterine activity. Analyzing may include determining FHR accelerations that increase >10 bpm above the precontraction FHR baseline for at least 15 seconds, and determining FHR decelerations that have FHR values less than the precontraction FHR rate during the contractions and / or after the contraction ends. Analyzing may include classifying FHR decelerations as early, variable, or late depending on the depth of the deceleration, its timing of onset and return to baseline FHR, symmetry, and repetitiveness. Analyzing may include determining the rates of change of deacclerations and / or accelerations, such as velocity and / or acceleration.

[0021] In accordance with further related embodiments of the invention, the method may be, at least in part, computer-implemented. Analyzing includes using a neural network. The neural network may be trained using predetermined fetal heart rate (FHR) and uterine activity data along with known occurrence of markers. The training may be suspended after a predetermined level of accuracy and / or precision is achieved.

[0022] In accordance with an embodiment of the invention, a system for treating a mother intrapartum is provided. The system includes sensors for obtaining fetal heart rate (FHR) and uterine activity data, including contraction data, from a pregnant mother intrapartum. A processor is configured to receive the fetal heart rate (FHR) and uterine activity data. The processor is further configured to determine contraction data. The contraction data may include a plurality of time periods, for example at least four time periods including: a first time period defined as predefined time before the contraction begins to the onset of the contraction; a second time period defined as duration of the contraction, a third time period defined as a predefined period after the contraction ends, and a fourth time period defined as any additional time after the end of the third time period till a predetermined time before the next contraction. The processor is further configured toanalyzed the FHR data and contraction data, within each of the four time periods, to determine occurrence of a marker selected from the markers consisting of Fetal Evaluation, Fetal Intervention, Fetal Operation, and Fetal Morbidity. A recommended intervention is provided by the processor based on the occurrence of the marker. The recommendation may be, for example, shown on a display.

[0023] In accordance with related embodiments of the invention, the recommended intervention may be selected from one or more of: no intervention required, improving tracing of the FHR pattern and / or contraction pattern by adjusting or switching the measuring equipment, desisting from pushing and / or desisting from inducing labor, reducing uterine activity by reducing Pitocin and / or administering uterine muscle relaxant, and delivery.

[0024] In accordance with further related embodiments of the invention, the recommended intervention for Fetal Evaluation (FE) may include: examine cervix to determine cervical dilation effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis; perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE); perform artificial rupture of membranes (AROM); reducing or stopping administration of picotin; place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion;if in 2nd stage (pushing), allow pushing only every other contraction, but if nonengaged do not allow push with each contraction; perform abdominal ultrasound to determine present fetal presentation; begin discussion of informed consent for operative vaginal delivery and begin consent process; inform labor staff of the potential need to perform operative intervention to help assemble the 2nd OR team; and / or evaluate for chori oamnionitis if fetal tachycardia or s / s chori oamnioitis with administration of antibiotics if IAI present or suspected.

[0025] In accordance with still further related embodiments of the invention, the recommended intervention for Fetal Intervention (FI) may include: examine cervix to determine cervical dilation, effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis; perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE); perform artificial rupture ofmembranes (AROM) if not already done; cease all pitocin induction / augmentation if being done and pushing efforts if in the 2nd stage of labor; place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion; administer 25 cc sc terbutaline if contractions > 6 / 10 minutes or hypertonic uterus; prepare staff for the potential of operative delivery; discuss and signing of operative delivery informed consent; full intrauterine resuscitation (IR) with cessation of all induction agents including pitocin, cervadil and / or cytotec; maternal repositioning, infusion of IV fluids; provide oxygen by mask; mnioinfusion if variable FHR decelerations have not started; move patient to operating room if FI persists more than 10 minutes; and / or prepare and examine patient for possible operative dl by anesthesiologist.

[0026] In accordance with yet further related embodiments of the invention, the recommended intervention for Fetal Operation (FO) includes: prep patient for operative; induce anesthesia if FHR tracing has not improved within 10 minutes; perform operative delivery; fetal head cooling after delivery; continue FHR monitoring following delivery for 1 hour to observe for characteristics of fetal tachycardia and absent variability (Category III tracing); perform c / section; and / or cord gases and placental pathology.

[0027] In accordance with more related embodiments of the invention, wherein analyzing, the processor may be configured to determining fetal features including baseline fetal heart rate (FHR), baseline variability, accelerations, decelerations, fetal recovery, suspected fetal acidosis, and / or fetal behavior. The analyzing may include determining uterine contraction parameters including baseline tone, frequency, amplitude, duration, intervals, rest-time, not interpretable, absent, rhythm, pushing, and / or uterine activity. The analyzing may include determining FHR accelerations that increase >10 bpm above the precontraction FHR baseline for at least 15 seconds, and determining FHR decelerations that have FHR values less than the precontraction FHR rate during the contractions and / or after the contraction ends. The analyzing may include classifying FHR decelerations as early, variable, or late depending on the depth of the deceleration, its timing of onset and return to baseline FHR, symmetry, and repetitiveness. The analyzing may include determining the rates of change of deacclerations and / or accelerations, such as velocity and / or acceleration.

[0028] In accordance with further related embodiments of the invention, the processor may include a neural network. The neural network may be trained using predetermined fetal heart rate (FHR) and uterine activity data along with known occurrence of markers. The training may be suspended after a predetermined level of accuracy and / or precision is achieved.

[0029] In accordance with another embodiment of the invention, a method for automated feature extraction is provided. The method includes acquiring a set of one or more non-digitized charts having one or more predetermined features. One or more markers is digitally assigned to the one or more predetermined features in the charts. The assigned marker-feature sets are digitally supplied to a supervised model. The model is statistically iterate over the assigned marker-feature sets. Model performance is automatically assessed one or more times during and / or after the statistical iteration process until a predetermined level of the model performance is achieved. The model is applied to one or more new sets of one or more charts to extract one or more non-assigned predetermined features.

[0030] In accordance with related embodiments of the invention, the set of one or more non-digitized charts may include a time-series set of one or more charts. Applying the model to one or more new sets of one or more charts may include applying the model to one or more digitized charts, signal data, and / or tabular data. The supervised model may include a convolution neural network model. The convolution neural network model may be configured to implement a basic single-shot detection deep learning approach. The markers may include one or more digital pixel coordinates corresponding to a location of the one or more features in the non-digitized charts. The charts may include one or more numeric arrays of pixel values with varying signal counts associated with the pixel values.

[0031] In accordance with further related embodiments of the invention, the method may further include splitting the assigned feature marker set in at least one training subset and at least one testing subset. The testing subset may comprise about 20% of the original assigned marker-feature set. Automatically assessing model performance may includeobserving model stability and / or minimization in its statistical loss function. The model may be configured to be integrated in a software application.

[0032] In accordance with another embodiment of the invention, a method for automated chart processing is provided. The method includes analyzing one or more time- series sets of one or more non-digitized charts of fetal heart rate (FHR) and / or concurrent maternal uterine contractions (UC) to digitally associate one or more markers with one or more predetermined fetal signatures in the charts. The associated marker- signature groups are automatically transmitted to an artificial intelligence model. The transmitted groups are used to perform supervised training of the artificial intelligence model. Accuracy and / or precision of the model is determined by comparing one or more predictions produced by the model to one or more previously unfamiliar to the model time-series sets of one or more charts of FHR and / or concurrent maternal UC with associated marker-signature groups. The model training is suspended after a predetermined level of accuracy and / or precision is achieved. The trained model is applied to automatically process one or more new time-series sets of one or more charts of FHR and / or concurrent maternal UC, having one or more un- associated predetermined fetal signatures.

[0033] In accordance with related embodiments of the invention, the artificial intelligence model may include a convolution neural network model configured to implement a basic single-shot detection deep learning approach. The supervised training of the artificial intelligence model may include statistically iterating over the transmitted groups to establish one or more associations between the markers and / or marker patterns in the FHR and / or UC with the one or more predetermined fetal signatures. The time-series sets may be obtained from non-digitized charts. Additional model training may include user feedback and / or provide additional training data. The fetal features may include fetal heart rate parameters including baseline fetal heart rate (stable, rising, falling, bradycardia, tachycardia, not defined, and / or uninterpretable), baseline variability (moderate, increased, decreased, and / or absent), accelerations (present, absent), decelerations (absent, early, late, prolonged, not identified), fetal recovery (normal, abnormal / overshoot, abnormal / delayed, abnormal / tachycardia), suspected fetal acidosis, and / or fetal behavior (normal - term, normal- preterm (later), abnormal, abnormal - suspected fetal injury). The fetal features may include uterine contraction parameters including baseline tone (normal range, increased / increasing, decreased / decreasing, not interpretable, and / or absent), frequency, amplitude, duration, intervals (between contractions, peak to peak), rest-time, not interpretable, absent, rhythm (no pattern, regular, irregular), pushing (absent, present), uterine activity (absent, uninterpretable, excessive, tachysystole, hypertonus, coupling / tripling, and / or rest time - decreased).

[0034] In accordance with still further related embodiments of the invention, the method may include additional processing of the one or more fetal signatures processed with the trained model. The method may further include generating a report and / or producing a notification for a user including an intervention recommendation (no intervention required, improve tracing of the FHR pattern (adjust the equipment, consider internal electrode, switch to an external device) and / or contraction pattern (adjust the equipment, consider an internal device), conservative actions (desist from pushing, desist from inducing labor, reduce uterine activity (reduce Pitocin, administer uterine muscle relaxant), consider delivery, adjust measuring equipment).

[0035] In accordance with any of the above-described embodiments, FE may be conditioned on contractions, late decelerations or variable decelerations or prolonged decelerations, and a prompt return to previously normal baseline rate and variability. FI may be conditioned on contractions, late decelerations or variable decelerations or prolonged decelerations, and a recovery to higher or lower rate and / or prolonged recovery overshoot after variable deceleration. FO may be conditioned on late decelerations or variable decelerations or prolonged decelerations, and blunted deceleration, overshoot with variable deceleration, and / or sinusoidal pattern during recovery. FM may be conditioned on agonal deceleration, usually prolonged or variable decelerations (but may be absent), and an unstable baseline. Adverse Modulations may include a conversion marker and / or Uterine Activity. A conversion marker while in normal, FE, or FI, may be defined as including the conditions of FO, and is marked by a sudden change to FO. Uterine activity such as abnormal contraction features may promote the class.Brief Description of the Drawings

[0036] The foregoing features of embodiments will be more readily understood by reference to the following detailed description, taken with reference to the accompanying drawings, in which:

[0037] Fig. 1 graphically depicts an example of a fetal heart rate (FHR) time series with a simultaneous capture of fetal heart signals (top) and maternal uterine contraction (UC) signals;

[0038] Fig. 2 graphically depicts an example of fetal signatures automatically identified in the fetus during labor, in accordance with an embodiment of the invention;

[0039] Fig. 3 schematically depicts a method for automated feature extraction, in accordance with an embodiment of the invention;

[0040] Fig. 4 schematically depicts a method for automated chart processing, according to an embodiment of the invention;

[0041] Fig. 5 is a digitally derived computer-generated labor patient FHR tracing from the beginning of the “Active Phase” of the First stage of labor, in accordance with an embodiment of the invention;

[0042] Fig. 6 is the same FHR tracing as Fig. 5 except with the addition of the computer generated shaded area under the curve from the apex of the acceleration to the FHR baseline rate of 136 bpm, in accordance with an embodiment of the invention;

[0043] Fig. 7 is a digitally derived computer-generated FHR tracing of the labor patient of Fig. 5 & 6, after 90 minutes in the active phase of the first stage of labor, in accordance with an embodiment of the invention;

[0044] Fig. 8 is the same as Fig. 7 except with the addition of the shaded area under the curve of the baseline (136 bpm) during the deceleration, in accordance with an embodiment of the invention;

[0045] Fig. 9 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5-8 after 2:04 hours of labor in the active phase of the first stage of labor after the performance of IR 30 minutes prior as the number of contractions in the 10 minutes has decreased to 2 contractions in the 10-minute period, in accordance with an embodiment of the invention;

[0046] Fig. 10 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5-9 after 4:08 hours of labor as the patient enters the second stage of labor, in accordance with an embodiment of the invention;

[0047] Fig. 11 is the same as Fig. 10 except with the FO marker shaded showing exaggeration of FHR variability during the FHR deceleration, in accordance with an embodiment of the invention;

[0048] Fig. 12 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5-11, after 4 hours and 30 minutes of labor, now 30 minutes into the 2ndstage of labor, and approximately 30 minutes after the FO marker first appeared, in accordance with an embodiment of the invention;

[0049] Fig. 13 is the same as Fig. 12 except with the FM marker highlighted in BLUE signifying the fetal tachycardia (>160 bpm) at 4: 13 which persisted till the fetus was delivered 15 minutes delivered while experiencing another FHR deceleration at 4:15, in accordance with an embodiment of the invention;

[0050] Fig. 14 shows a digitally derived computer-generated labor patient FHR tracing, in accordance with an embodiment of the invention;

[0051] Fig. 15 shows a tracing of the same patient as Fig. 14 further in time, showing a Conversion Marker signifying a sudden change to FO, in accordance with an embodiment of the invention;

[0052] Fig. 16 shows in more detail FE-FI-FO-FM classification management, in accordance with an embodiment of the invention; and

[0053] Fig. 17 providing an explanation key for Fig. 16.Detailed Description of Specific Embodiments

[0054] Illustrative embodiments of the invention relate to systems and methods for applying machine learning to feature recognition and processing in both digitized and non- digitized fetal tracings to establish mathematical and visual recognition patterns of the Fetal Heart Rate (FHR) tracing, including Fetal Heart Rate (FHR), Fetal Variability (FV), FHR accelerations (FA), FHR Decelerations (FD), along with uterine activity (UA), uterine contraction tachy systole and excessive uterine activity (EXU A). The method visually and mathematically determines threats to Fetal Reserve (FR) that may be unrelated to discernible changes in the FHR and to identify critical action points earlier than traditional methods of fetal monitoring interpretation for safeguarding the previously normal fetus while hastening intervention in the previously compromised fetus by timely evaluation and intervention by intrauterine resuscitation (IR) to avoid emergency operative delivery. The system and methodology advantageously uses an approach that is, termed FE-FI-FO-FM (pronounced Fee-Phi-Pho-Fumm):

[0055] FE (Fetal Evaluation) - As used in this description and the accompanying claims, the term “Fetal Evaluation (FE) shall mean, unless the context otherwise requires: the definition of the normally behaving fetus in the presence of mild changes in the FHR tracing but in the absence of any obvious hypoxic, ischemic, mechanical, or infectious threat.

[0056] FI (Fetal Intervention) - As used in this description and the accompanying claims, the term “Fetal Intervention (FI) shall mean, unless the context otherwise requires: the recognition of a hypoxic, ischemic, mechanical, or infectious threats to the fetus (ormother) that should promote remedial action in the expectation of recovery (such as, without limitation, Intrauterine Resuscitation (IR) with IV fluids, Oxygen, maternal position change, cessation of Pitocin infusion, amnioinfusion).

[0057] FO (Fetal Operation) - As used in this description and the accompanying claims, the term “Fetal Operation (FO) shall mean, unless the context otherwise requires: the recognition of a hypoxic, ischemic, mechanical, or infectious threat to the fetus (or mother) that should promote immediate intervention where recovery is improbable with conservative action (such as, the need to expedite vaginal or abdominal delivery- rescue).

[0058] FM (Fetal Morbidity) - As used in this description and the accompanying claims, the term “Fetal Morbidity (FM) shall mean, unless the context otherwise requires: the identification of likely neurological injury in the fetus. The may include, for example: A. a pre-existing injury (e.g., a pattern on admission), or B. the conversion from a normal fetal behavior and responsive neurologic status to identifiable fetal injury with implications of neonatal neuroprotection.

[0059] The Primary objective of FE-FI-FO-FM is to alert the clinician to the need to be proactive in the assessment and treatment of laboring patients in order to keep the fetus (and mother) out of harm’s way in the first place by safely conducting labor and avoiding the need for emergency intervention. The prospective FE-FI-FO-FM methodology / system / computerized algorithm (e.g., using artificial intelligence) assists the Obstetrician in identifying fetal status alterations in order to timely predict and prevent the onset of irreversible fetal neurological injury. Prior conventional methodologies / procedures use an endpoint of the classification of FHR characteristics to intervene, however, only after the onset of irreversible fetal injury. Embodiments described herein represent a major advancement in the logical identification of prospective injury in order to prevent the injury from occurring, which may use computer assistance to identify new markers which are harbingers of the development of irreversible fetal brain damage if intervention is not immediately undertaken. This methodology champions prospective prevention by timely intervention rather than reactive management to fetal injury that has already occurred.

[0060] Generally, there is a standard set of FHR features that may also be associated with uterine contractions (UCs), which are being monitored in conjunction with FHR. Typically, UCs provoke responses in the fetal heart rate pattern, including anticipatory accelerations, or decelerations. Evaluating the timing and the pattern of recovery helps to define the robustness and propriety of the response, and the mechanism of any compromise. It permits differentiation of oxygen impairment related to insufficient oxygen availability and / or interruption of blood flow in the umbilical cord or the cerebral circulation. UC patterns can be used as relevant features informing the assessment of the progress of labor and future threats to the fetus. For example, if a fetus is unable to recover to a normal heart rate baseline post contraction, one may infer that the fetus is attempting to compensate for a problem elicited by the stress of the contraction. Signatures that define the need for attention to the fetus can be designated by one or more changes in amplitude, period, irregularity, regularity, and response duration to the contraction. Such fetal signatures can be dependent or independent of the contraction.

[0061] Illustrative embodiments of the invention allow for the mathematical calculation of fetal heart rate monitoring and uterine contraction characteristics that can only be determined mathematically to better quantify the altered fetal condition during periods of uterine contraction and relaxation throughout the stages and phases of labor. This may be accomplished by determining from the contraction data, a plurality of time periods, and analyzing the FHR data and contraction data, within each of the plurality of time periods, to determine occurrence of a marker (i.e., Fetal Evaluation (FE), Fetal Intervention (FI), Fetal Operation (FO), and Fetal Morbidity (FM)). The novelty of the computerized analysis is to divide the contraction and relaxation periods into separate time events and compare the subtle changes that occur prior to, during, after the contraction occurs so that pathophysiological changes in the control of the parasympathetic nervious system (vagus nerve control of the fetal heart) can earlier predict impending neurological injury than the naked eye by utilizing computer velocity of change and acceleration coefficients of change to earlier predict disfunction of the vagus nerve control of the fetal heart rate.

[0062] The time periods may consist of, for example and without limitation, four distinct time periods. The first time period may be a predefined time before the contraction begins to the onset of a contraction (i.e., the precontraction time period). This predefined time may be, without limitation, between 15 and 35 seconds (e.g., 30 seconds prior to the onset of a contraction). This first time period may be akin to a 100 yard dash runner at the starting line - the runner's heart rate may be slightly elevated due to anticipation of the starting gun (or in the case of the fetus, the starting of the contraction). The second time period is defined as the duration of the contraction (i.e., the contraction time period). The second period is akin to the stress and elevated heart rate of the runner while running the race. The third time period is a predefined period after the contraction ends (i.e., the post contraction time period). This predefined time may be, without limitation, between 30 seconds to a 1 minute 10 seconds (e.g., 1 minute). This third period is akin to when the runner just finishes the race. Lastly, the fourth time period is defined as any additional time after the end of the third time period till a predefined time before the next contraction (i.e., before the first time period). The predefined time before the next contraction may be, without limitation, between 25 and 35 seconds before the next contraction (e.g., 30 seconds). This fourth period is akin to the extra time (e.g., additional recovery and / or relaxation time) the runner has before he needs to run another race The FHR data and contraction data is analyzed within each of the four above-described time periods so as to provide an assessment of the progress in labor. The four time periods listed above may be determined by analyzing baseline FHR data taken early in labor, for example, during 10 minutes of early normal intrapartum. Furthermore, the time periods may be iteratively determined - for example, they may be periodically determined throughout labor.

[0063] In various embodiments of the invention, a further evaluation of the Neonatal Heart Rate (NHR) pattern in the immediate newborn period (approximately 1 hour after birth) may be performed to identify the risk of permanent fetal neurologic injury and the risk of Autism Syndrome (AS) and Hypoxic-Ischemic Encephalopathy (HIE) from reduction in Fetal Reserve during labor and the need for Neonatal Cooling (NC).

[0064] By using a combination of mathematical and visual patterns of progress in labor, uterine activity and the response (if any) of the fetus, one can also elucidate preventative approaches to better safeguard the fetal and mother from future stresses and emergency operative intervention. This approach is advantageous to current methodology, both clinical and computer-based.

[0065] Furthermore, this approach to labor surveillance (FE-FI-FO-FM) offers unique short-term approaches to the evaluation of the quality of obstetrical care (the reduction in emergency interventions - a process evaluation) as well as standard clinical outcome evaluations (including long-term follow up) which can be critically compared to current fetal monitoring evaluation methodology and to determine needs for education and training.

[0066] The FHR UC monitoring data may be obtained real-time digitally via direct input from the fetal monitor or from FHR data files previous collected. However, most of the tracing data of previously delivered babies can only be obtained by reviewing paper printouts of the tracings to obtain images of the tracing which then can be scanned and digitized. The FHR tracing data can be analyzed from either source (digital or tracing image) to obtain similar interpretation and characterization of aspects of the FHR and UC tracings. Once the FHR mages have been digitized, they will be analyzed in similar fashion to the original digitized data collected real-time or in stored files that come directly from the Fetal Heart Rate monitor.

[0067] In various embodiments, the present invention embraces applying machine learning methods for supervised model training for automated feature extraction from charts. The present invention is based on a concept of treating one or more non-vectorized images (such as graphs, plots, and / or other graphically or tabularly represented data) as traditional ‘images’, similar to digital photographs, and applying supervised machine learning to extract and / or process features to train an artificial intelligence model. In this approach, an image (herein, also referred to as chart) is supplied to a convolution neural network (CNN) model. The image may be represented, without limitation, as one or more numeric arrays of pixelvalues with varying signal counts associated with the pixel content (also referred to as data and / or values). The pixel content may be dictated by the amount of red, green, blue or other spectral bands that the pixel may receive, and is an integer number in one or more dimensions. The CNN is represented as a set of algorithmic layers into which the numeric pixel data is sent. It may consist of a series of convolutional layers, nonlinear layers, pooling layers, and / or fully connected layers. Each such layer may be considered an individual set of equations, where the output of one equation becomes the input to another. The CNN eliminates the need for manual feature extraction, as the features are acquired through the passing of the pixel data to one or more other layers, and correlations are extracted and weighted as a consequence of the layer transitions.

[0068] For a supervised approach, the model is provided one or more labeled examples of the important features (labels, markers, and / or training data) for the CNN to generate correlations among the images with the features. The CNN then adapts its algorithm to try to predict the relevant features through statistical iteration of the features extracted from one or more layers in the CNN. For example, a Single Shot Detector (SSD) algorithm can be applied. It utilizes a standard CNN network (e.g., VGG-16) with an additional set of convolution layers to identify discrete locations of one or more features in one or more images. Alternatively or additionally, RCNN, LSTM, RNN, support vector machine, random forest, instance segmentation, image classification techniques, and / or other deep learning algorithms and / or other machine learning techniques can be applied. Through a single pass in the CNN, the weighted correlations meant to describe the relevant features may be tested against a truth data (validation and / or test data), separate from training data. The goal of this statistically iterative operation is to minimize a loss function between the predicted correlations and the truth values through adaptively updating the weights of the predicted function. The process of adjusting the weights can continue until a minimum statistical loss is obtained. The output model and weights may then be used for inference against a new data set to extract similar relevant information. The new data set may be supplied to the trained model in a format different from the format of the original training / testing images. For example, the new data set may be supplied in the format of digitized charts, tabularly represented data, a signal received from one or more devices, etc. In other words, once themodel has been trained, it can be configured to work on similar features provided in the same and / or other data formats. Such an approach allows the model to identify the one or more features of interest, and also the location of those features in the chart(s). This location can be correlated with a time and / or other dependent variable(s) within the chart and / or a set of charts.

[0069] In an embodiment described herein, the above-described approach may be applied to the identification of fetal signatures requiring attention, such as anomalous heartbeat in response to contractions or other provocations (e.g. separation of the placenta from the uterus, prolapse of the umbilical cord), using an archive of medical features gathered by expert physicians over many years. Fig. 2 graphically depicts an example of fetal signatures requiring attention (specifically, late decelerations) automatically identified in the fetus during labor, in accordance with an embodiment of the invention. These signatures are outlined with the black dashed boxes. Compared to Fig. 1, there are notable trough features in this fetal heart signal. With every contraction, the fetus’ heart rate is altered at the time of contraction, to which the fetus recovers from the event and returns its heart rate to a normal baseline. The repetition and duration of the troughs depict significant stress on the fetus, which may still possess sufficient reserve to recover to a normal baseline heart rate. Often this may be a result of diminished oxygen availability from diminished uterine blood flow and / or oxygen reduction in the mother that have occurred during the labor and are reflected by the response to the stress of the contractions.

[0070] The approach may be further configured to be implemented in a pro-active fetal surveillance system. The system may be used to identify critical features in real-time. For example, this can provide information to a user (for example, a medical professional, expectant mother), indicating a need for adjusting surveillance devices and / or procedures during labor to obtain an interpretable tracing. The system may be configured to be implemented in a stationary, portable, or partially portable device that may include, or hooks up to, a processor. Examples of such devices include a mobilenet application, medical devices used for FHR monitoring, and wearable technology such as a belt with a sensor for registering FHR and / or maternal UC.

[0071] This system and methodology may be configured to be used for medical-legal applications to assist in testimony associated with malpractice lawsuits. Additionally or alternatively, the system and methodology may be used in the field of hospital administration, such as for assessing doctor performance and / or fetal and / or maternal health risk. Additionally or alternatively, the system and methodology may be used in-situ with a patient during pregnancy and / or labor to advise the user as to the fetus’ or mother’s health. The system and methodology can also be used for training and / or educational purposes.

[0072] Fig. 3 schematically depicts a method 100 for automated feature extraction, according to an embodiment of the invention. At 102, a set of one or more non-digitized charts having one or more predetermined features is acquired. At 104, one or more markers are digitally assigned to the one or more predetermined features in the charts. At 106, the assigned marker-feature sets are digitally supplied to a supervised model. At 108, the model is used to statistically iterate over the assigned marker-feature sets. At 110, model performance is automatically assessed one or more times during and / or after the statistical iteration process until a predetermined level of the model performance is achieved. At 112, the model is applied to one or more new sets of one or more charts to extract one or more non-assigned predetermined features.

[0073] In various embodiments, the charts may comprise one or more numeric arrays of pixel values with varying signal counts associated with the pixel values. The set of one or more nondigitized charts may include a time-series set of one or more charts. In an embodiment, the supervised model may include a convolution neural network (CNN) model. Additionally, the CNN model may be configured to implement a basic single-shot detection deep learning approach. Additionally or alternatively, RCNN, LSTM, support vector machine, random forest, instance segmentation, image classification techniques, and / or other deep learning algorithms and / or other machine learning techniques may be applied. Applying the model to one or more new sets of one or more charts at 112 may includee applying the model to one or more digitized charts, signal data, and / or tabular data. Automaticallyassessing model performance may include observing model stability and / or minimization in its statistical loss function. In an embodiment, the markers assigned at 104 may include one or more digital pixel coordinates corresponding to a location of the one or more features in the non-digitized charts. The method 100 may further include splitting the assigned feature- marker set in at least one training subset and at least one testing subset. Additionally, the testing 290 subset may comprise about 20% of the original assigned marker-feature set.[00741 Invarious embodiments, the model may be configured to be integrated in a software application. The software application may be used in a variety of systems and devices. For example, the application may be configured to be used with a cell phone, a computer, specialized medical equipment, and / or wearable piece of technology intended for personal and / or professional use. Additionally or alternatively, the model output may be combined with clinical data from an electronic medical record (EMR), providing a more complete assessment of the labor 300 than is currently available.

[0075] Fig. 4 schematically depicts a method 200 for automated chart processing, in accordance with an embodiment of the invention. At 202, one or more time-series sets of one or more non-digitized charts of fetal heart rate (FHR) and / or concurrent maternal uterine contractions (UC) may be analyzed to digitally associate one or more markers with one or more predetermined fetal signatures in the charts. At 204, the associated marker-signature groups are automatically transmitted to an artificial intelligence model. At 206, the transmitted groups are used to perform supervised training of the artificial intelligence model. At 208, accuracy and / or precision of the model are determined by comparing one or more 310 predictions produced by the model to one or more previously unfamiliar to the model 11 time-series sets of one or more charts of FHR and / or concurrent maternal UC with associated marker-signature groups. At 210, the model training is suspended after a predetermined level of accuracy and / or precision is achieved. At 212, the trained model is applied to automatically process one or more new time-series sets of one or more 315 charts of FHR and / or concurrent maternal UC, having one or more un-associated predetermined fetal signatures.

[0076] In various embodiments, the artificial intelligence model may include a convolution neural network model (CNN). The CNN model may be further configured to implement a basic single shot detection deep learning approach. Additionally or alternatively, the artificial intelligence model may include RCNN, LSTM, support vector machine, random forest, instance segmentation, image classification techniques, and / or other deep learning algorithms and / or other machine learning techniques. The supervised training of the artificial intelligence model at 206 may include statistically iterating over the transmitted 325 groups to establish one or more associations between the markers and / or marker patterns in the FHR and / or UC with the one or more predetermined fetal signatures. The method 200 may further include additional model training to include user feedback and / or provide additional training data. In an embodiment, the time-series sets can be obtained from non-digitized charts (for example, pdf files).

[0077] The fetal features may include fetal heart rate parameters including baseline fetal heart rate (stable, rising, falling, bradycardia, tachycardia, not defined, and / or uninterpretable), baseline variability (moderate, increased, decreased, and / or absent), accelerations (present, absent), decelerations (absent, early, late, prolonged, 335 not identified), fetal recovery (normal, abnormal / overshoot, abnormal / delayed, abnormal / tachycardia), suspected fetal acidosis, and / or fetal behavior (normal - term, normal - preterm (later), abnormal, abnormal - suspected fetal injury).

[0078] Additionally or alternatively, the fetal features may include uterine contraction parameters including baseline tone (normal range, increased / increasing, decreased / decreasing, not interpretable, and / or absent), frequency, amplitude, duration, intervals (between contractions, peak to peak), rest-time, not interpretable, absent, rhythm (no pattern, regular, irregular), pushing (absent, present), uterine activity (absent, uninterpretable, excessive, tachysystole, hypertonus, coupling / tripling, and / or rest time - decreased).

[0079] In various embodiments, the method 200 may further include additional processing of the one or more fetal signatures processed with the trained model. Additionally, the method 200 may further include generating a report and / or producing a notification for a user including an intervention recommendation (no intervention required, improve tracing of the FHR pattern (adjust the equipment, consider internal electrode, switch to an external 350 device) and / or contraction pattern (adjust the equipment, consider internal device), conservative actions (desist from pushing, desist from inducing labor, reduce uterine activity (reduce Pitocin, administer uterine muscle relaxant), consider delivery, adjust measuring equipment). Additionally or alternatively, the report may include a timestamp of the detected features. Future timestamps may also be predicted for one or more features. For detections made for one or more components in the chart (e.g., contractions), concurrent features can be extracted for other components (e.g., FHR). Predictions can include timestamps of feature detections, time duration for individual features, and / or a difference between consecutive features, and / or relative feature identifications per image in the document.

[0080] Fig. 5 is a digitally derived computer-generated labor patient FHR tracing from the beginning of the “Active Phase” of the First stage of labor, in accordance with an embodiment of the invention. The data may be obtained in real-time or in electronic data files, which may be the result of scanned charts. The FHR data is typically an integer number between 30-240 and represents beats per minute (BPM). The contraction data is also typically an integer number between 0-100 mmHg (pressure measurement). The digitized FHR monitoring data is plotted in a graph as in Fig. 2. As the digitized data becomes available in real-time or from the stored data file, the uterine contraction baseline tone (BT) shall be calculated and when the tone increases, the time and duration of the contraction is assessed to determine the beginning and ending of contractions when the BT data returns to precontraction UC level.

[0081] The bottom aspect of Fig. 5 depicts the uterine contraction baseline uterine tone and three (1,2,3) contractions in the first 10 minutes with the vertical shading representing the duration of the contractions. The upper aspect of Fig. 5 is the Fetal heartrate tracing with representative Pre, During, Post, and Relaxation divisions of the uterine contraction affect on the fetal response to the contraction and degree of fetal reserve. Depicted is the FHR baseline (mean) and the FHR variability calculations (mean, and range). Above contraction #2, a typical FHR acceleration is identified, and there are no FHR decelerations noted in this ten-minute segment.

[0082] Fig. 6 is the same FHR tracing as Fig. 5 except with the addition of the computer generated shaded area under the curve from the apex of the acceleration to the FHR baseline rate of 136 bpm, in accordance with an embodiment of the invention. The shaded area will be utilized to calculate the area of fetal reserve (FHR acceleration) and its velocity and acceleration coefficients of change from baseline in relation to fetal reserve.

[0083] Fig. 7 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5 & 6, after 90 minutes in the active phase of the first stage of labor, in accordance with an embodiment of the invention. Again, the upper and lower aspects of the Fig. 7 depict the FHR tracing and uterine activity aspects, respectively. A Pre-FHR Acceleration is noted prior to the onset of the first FHR tracing abnormality termed a “prolonged FHR deceleration” and labeled FE (Fetal Evaluation) after the onset of the “coupled contraction (#1 A) (abnormal) which continues for over 5 minutes without return to the baseline UC tone of 30, the period of coupling (between 1 A+1B) without return to baseline uterine tone (45 mm Hg) and its eventual relaxation after 5.5 minutes to baseline uterine tone by 1 :37 and return to the FHR baseline of 136 at the end of the FHR tracing. In addition to the four FHR zones (Pre, During, Post, and relaxation) of the contraction, a cross- hatched vertical area lasting 80 seconds is depicted representing marked increase in the FHR and FHR variability during the coupled contraction (FI) which is the first marker for the need for immediate Fetal intervention which traditionally includes IV fluids, cessation of oxytocin, oxygen by face mask, maternal position change, and amnioinfusion if possible). Note that intervention at this time does not include cesarean delivery, rather improving the intrauterine environment with Intrauterine resuscitation of the fetus.

[0084] Fig. 8 is the same as Fig. 7 except with a shaded area under the curve of the baseline (136 bpm) during the deceleration, in accordance with an embodiment of the invention. Again, velocity and acceleration coefficients are calculated from the abrupt and significant FHR Variability that exceeds 25 bpm in the cross-hatched zone at 1 : 14:35 of the tracing. Mathematical calculation of uterine activity and relaxation time between contractions will depict the relative time of cessation of blood flow to the fetus and recovery of blood flow.

[0085] Fig. 9 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5-8 after 2:04 hours of labor in the active phase of the first stage of labor after the performance of IR 30 minutes prior as the number of contractions in the 10 minutes has decreased to 2 contractions in the 10-minute period, in accordance with an embodiment of the invention. Of note is the baseline FHR of 136 bpm with an acceleration during the contraction period that failed to react “Reactivity” because it failed to increase 10 bpm over baseline, mild decreased FHR variability, and the onset of a “late” FHR deceleration at 2:09 labelled “Late FHR deceleration” and also noting the FI (Fetal Intervention) label indicative of the need for further Intrauterine Resuscitation (IR).

[0086] Fig. 10 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5-9 after 4:08 hours of labor as the patient enters the second stage of labor, in accordance with an embodiment of the invention. Again the uterine activity aspect of Fig. 10 depicts a coupled contraction (1A and IB) and contractions 2 and 3, with shaded areas depicting the duration of uterine activity. Again, the lack of a FHR acceleration failing to meet “Reactivity” criteria (10 bpm for 15 seconds) is seen with a baseline FHR 137 bpm and the onset of the Variable FHR deceleration with the FHR Variability Marker identifying a fetal condition that requires fetal operation (FO) because the fetal reserve and FHR tracing have evolved into a hypoxic / acidotic period with decreased FHR baseline variability, deep variable FHR decelerations that do not recover at the end of the FHR contractions and marked FHR variability during the uterine contraction. FO depicts the presence of a fetal reserve that is inadequate to safely labor the mother and fetus thereby requiring immediate intervention (IR) and operative abdominal or vaginal delivery.

[0087] Fig. 11 is the same as Fig. 10 except with the FO marker highlighted in vertical shaded exaggeration of FHR variability during the FHR deceleration, in accordance with an embodiment of the invention. Again, FHR baseline rate mean, FHR variability, FHR acceleration, and deceleration timing, during, velocity and acceleration coefficients are calculated that depict the presence of an ominous decrease in fetal reserve requiring patient delivery.

[0088] Fig. 12 is a digitally derived computer-generated FHR tracing of the labor patient of Figs. 5-11, after 4 hours and 30 minutes of labor, now 30 minutes into the 2ndstage of labor, and approximately 30 minutes after the FO marker first appeared, in accordance with an embodiment of the invention: Management at this point included 1R (cessation of all pushing efforts, maternal position change, Oxygen by mask, and cessation of all Pitocin infusion) following the onset of the prolonged FHR Variable deceleration, onset of fetal arrhythmia, and overshoot of the FHR to a Fetal Tachysystole > 180 bpm, the FM Marker (Fetal Conversion and Irreversible Fetal Injury) first appeared.

[0089] Fig. 13 is the same as Fig. 12 except with the FM marker highlighted in a light shade signifying the fetal tachycardia (>160 bpm) at 4: 13 which persisted till the fetus was delivered 15 minutes delivered while experiencing another FHR deceleration at 4: 15, in accordance with an embodiment of the invention; The neonatal outcome included low Neonatal Apgar Scores, Metabolic acidosis, Autism Syndrome and mild HIE with characteristic MRI (magnetic resonance imaging) findings. It is to be emphasized that the FE-FI-FO-FM markers all occurred prior to the onset of the appearance of the ACOG Category System Marker currently used today as the onset of the need to consider delivery of the fetus. In this case, a TRUE CATEGORY 777 tracing never occurred.

[0090] Figs. 14 and 15 show a digitally derived computer-generated labor patient FHR tracing that results in a Conversion Marker signifying a sudden change to FO, in accordance with an embodiment of the invention. The computer calculations, some of which are summarized in Table One, include, without limitation, the following:

[0091] Time periods for each contraction (total of 4):1) Precontraction2) Contraction3) Post Contraction4) Recovery Period

[0092] FHR ANALYSIS OF: a. FHR baseline (mean rate, range, std deviation) i. Tachycardia (>160 bpm) ii. Bradycardia (<110 bpm)

[0093] b. FHR baseline variability (mean, absolute value mean, range, std deviation) i. Absent (0-1 beats) ii. Minimal (1-5 beats) iii. Moderate (6-24 beats) iv. Marked (>24 beats) v. CM / FO marker

[0094] c FHR Accelerations (mean rate, range, std deviation, duration, coefficients of velocity and acceleration change) i. Reactive (2 or more in 20 minutes) ii. Nonreactive (<2 in 20 minutes)

[0095] d. FHR Decelerations (mean rate, range, std deviation, onset, duration, uniformity, repetitive, coefficients of velocity and acceleration change) i. Early ii. Variable iii. Late iv. Prolonged

[0096] Uterine Activity Calculations1. Number of contractions in 10 minutes2. Tone (baseline)3. Height of contraction (mmHg)4. Montivideo Units (peak - tone for 10 minute period)5. Total time in contractions for 10 minutes6. Total Relaxation time in 10 minutes (10 minute minus Contraction time7. Tachysystole (>5 contractions in 10 minutes)

[0097] All of the FHR and Uterine Contraction calculations were, without limitation, ongoing with summaries for each contraction, previous 10 minutes of tracing, previous 30 minutes of tracing, and previous 1 hour of tracing. The FE-FI-FO-FM Calculations were, without limitation, analyzed with comparison to1. Pre-2003 FHR monitoring analysis standards2. ACOG Category System3. Fetal Reserve Index Analysis

[0098] Fig. 14 shows ten minutes of FHR monitoring, in which there are three contractions. Fig. 15 shows, at roughly 1 : 14:45 a coupled contraction (labelled UC 21) occurs, that follows a period of blunted deceleration, continuing with overshoot with variable deacceleration irrespective of amplitude, including a sinusoidal pattern, which represents a Conversion Marker leading directly to FO.

[0099] Fig. 16 shows in more detail FE-FI-FO-FM classification management, in accordance with various embodiments of the invention, with Fig. 17 providing an explanation key. In Fig. 16, the FE-FI-FO-FM columns include conditions for occurrence for both baseline FHR features and decelerative patterns. Arrows show that that movement between the columns can be left-right (an arrow pointing left and right) or only in one direction (an arrow pointing in only one direction). For example, when in FE, the fetus can recover to normal (likely to happen), when in FI, the fetus can recover to FE (possible), when in FO one can recover to FI (unlikely). When in FM, the fetus is injured.

[0100] In addition to the columns showing associated conditions, there are two adverse modulations. The first is labelled uterine activity and is defined in Fig. 17 with regard to normal and excessive uterine activity. As described in the notes , an abnormal orexcessive contraction feature will cause promotion within FE-FI-FO-FM. The second is a conversion marker in which a fetus can transition to FO regardless of prior state if the conditions for FO are met.

[0101] Embodiments of the present invention may be embodied in many different forms, including, but in no way limited to, computer program logic for use with a processor (e.g., a microprocessor, microcontroller, digital signal processor, or general purpose computer), programmable logic for use with a programmable logic device (e.g., a Field Programmable Gate Array (FPGA) or other PLD), discrete components, integrated circuitry (e.g., an Application Specific Integrated Circuit (ASIC)), or any other means including any combination thereof.

[0102] Computer program logic implementing all or part of the functionality previously described herein may be embodied in various forms, including, but in no way limited to, a source code form, a computer executable form, and various intermediate forms (e.g., forms generated by an assembler, compiler, networker, or locator.) Source code may include a series of computer program instructions implemented in any of various programming languages (e.g., an object code, an assembly language, or a high-level language such as Fortran, C, C++, JAVA, or HTML) for use with various operating systems or operating environments. The source code may define and use various data structures and communication messages. The source code may be in a computer executable form (e.g., via an interpreter), or the source code may be converted (e.g., via a translator, assembler, or compiler) into a computer executable form.

[0103] The computer program may be fixed in any form (e.g., source code form, computer executable form, or an intermediate form) either permanently or transitorily in a tangible storage medium, such as a semiconductor memory device (e.g., a RAM, ROM, PROM, EEPROM, or Flash-Programmable RAM), a magnetic memory device (e.g., a diskette or fixed disk), an optical memory device (e.g., a CD-ROM), a PC card (e.g., PCMCIA card), or other memory device. The computer program may be fixed in any form in a signal that is transmittable to a computer using any of various communicationtechnologies, including, but in no way limited to, analog technologies, digital technologies, optical technologies, wireless technologies, networking technologies, and internetworking technologies. The computer program may be distributed in any form as a removable storage medium with accompanying printed or electronic documentation (e.g., shrink wrapped software or a magnetic tape), preloaded with a computer system (e.g., on system ROM or fixed disk), or distributed from a server or electronic bulletin board over the communication system (e.g., the Internet or World Wide Web).

[0104] Hardware logic (including programmable logic for use with a programmable logic device) implementing all or part of the functionality previously described herein may be designed using traditional manual methods, or may be designed, captured, simulated, or documented electronically using various tools, such as Computer Aided Design (CAD), a hardware description language (e.g., VHDL or AHDL), or a PLD programming language (e g., PALASM, ABEL, or CUPL).

[0105] The embodiments of the invention described above are intended to be merely exemplary; numerous variations and modifications will be apparent to those skilled in the art. All such variations and modifications are intended to be within the scope of the present invention as defined in any appended claims. The use of the term “and / or” includes any and all combinations of one or more of the associated listed items. The figures are schematic representations and so are not necessarily drawn to scale. Unless otherwise noted, specific terms have been used in a generic and descriptive sense and not for purposes of limitation.Table One

Claims

What is claimed is:

1. A method of treating a mother intrapartum, the method comprising: obtaining fetal heart rate (FHR) and uterine activity data, including contraction data, from a pregnant mother intrapartum; determining from the contraction data, for each contraction, a plurality of time periods: analyzing the FHR data and contraction data, within each of the plurality of time periods, to determine occurrence of a marker selected from the markers consisting of Fetal Evaluation (FE), Fetal Intervention (FI), Fetal Operation (FO), and Fetal Morbidity (FM); and providing a recommended intervention based on the occurrence of the marker.

2. The method according to claim 1, wherein the plurality of time periods includes: a first time period defined as predefined time before the contraction begins to the onset of the contraction, a second time period defined as duration of the contraction, a third time period defined as a predefined period after the contraction ends, and a fourth time period defined as any additional time after the end of the third time period till a predetermined time before the next contraction;3. The method according to claim 1, further comprising performing the recommended intervention.

4. The method according to claim 1, wherein the recommended intervention is selected from one or more of: no intervention required, improving tracing of the FHR pattern and / or contraction pattern by adjusting or switching the measuring equipment, desisting from pushing and / or desisting from inducing labor, reducing uterine activity by reducing Pitocin and / or administering uterine muscle relaxant, and delivery.

5. The method according to claim 1, wherein the recommended intervention for FetalEvaluation (FE) includes: examine cervix to determine cervical dilation, effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis, perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE), perform artificial rupture of membranes (AROM), reducing or stopping administration of picotin, place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion, if in 2nd stage (pushing), allow pushing only every other contraction, but if nonengaged do not allow push with each contraction, perform abdominal ultrasound to determine present fetal presentation, begin discussion of informed consent for operative vaginal delivery and begin consent process, inform labor staff of the potential need to perform operative intervention to help assemble the 2nd OR team, and / or evaluate for chorioamnionitis if fetal tachycardia or s / s chori oamnioitis with administration of antibiotics if IAI present or suspected.

6. The method according to claim 1, wherein the recommended intervention for Fetal Intervention (FI) includes: examine cervix to determine cervical dilation, effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis, perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE), perform artificial rupture of membranes (AROM) if not already done, cease all pitocin induction / augmentation if being done and pushing efforts if in the 2nd stage of labor, place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion, administer 25 cc sc terbutaline if contractions > 6 / 10 minutes or hypertonic uterus, prepare staff for the potential of operative delivery,discuss and signing of operative delivery informed consent, full intrauterine resuscitation (IR) with cessation of all induction agents including pitocin, cervadil and / or cytotec, maternal repositioning, infusion of IV fluids, provide oxygen by mask, amnioinfusion if variable FHR decelerations have not started, move patient to operating room if FI persists more than 10 minutes, and / or prepare and examine patient for possible operative dl by anesthesiologist.

7. The method according to claim 1, wherein the recommended intervention for Fetal Operation (FO) includes: prep patient for operative, induce anesthesia if FHR tracing has not improved within 10 minutes, perform operative delivery, fetal head cooling after delivery, continue FHR monitoring following delivery for 1 hour to observe for characteristics of fetal tachycardia and absent variability (Category III tracing), perform c / section, and / or cord gases and placental pathology.

8. The method according to claim 1, wherein the recommended intervention for Fetal Morbidity (FM) includes: full neonatal resuscitation with Fetal scalp electrode on the fetus after delivery, emergency delivery, brain cooling, and / or cord gases and placental pathology.

9. The method according to claim 1, wherein analyzing includes determining fetal features including baseline fetal heart rate (FHR), baseline variability, accelerations, decelerations, fetal recovery, suspected fetal acidosis, and / or fetal behavior.

10. The method according to Claim 1, wherein analyzing include determining uterine contraction parameters including baseline tone, frequency, amplitude, duration, intervals, rest-time, not interpretable, absent, rhythm, pushing, and / or uterine activity.

11. The method according to claim 1, wherein analyzing includes determining FHR accelerations that increase >10 bpm above the precontraction FHR baseline for at least 15 seconds, and determining FHR decelerations that have FHR values less than the precontraction FHR rate during the contractions and / or after the contraction ends.

12. The method according to claim 1, wherein analyzing includes classifying FHR decelerations as early, variable, or late depending on the depth of the deceleration, its timing of onset and return to baseline FHR, symmetry, and repetitiveness.

13. The method according to claim 1, wherein analyzing includes determining the rates of change of deacclerations and / or accelerations.

14. The method according to claim 13, wherein the rates of change include velocity and / or acceleration.

15. The method according to claim 1, wherein the method is, at least in part, computer- implemented.

16. The method according to claim 15, wherein analyzing includes using a neural network.

17. The method according to claim 16, further comprising training the neural network using predetermined fetal heart rate (FHR) and uterine activity data along with known occurrence of markers.

18. The method according to claim 17, further comprising suspending the training after a predetermined level of accuracy and / or precision is achieved.

19. A system for treating a mother intrapartum, the system comprising: sensors for obtaining fetal heart rate (FHR) and uterine activity data, including contraction data, from a pregnant mother intrapartum; a processor configured to: receive the fetal heart rate (FHR) and uterine activity data; determine from the contraction data, for each contraction, a plurality of time periods; analyze the FHR data and contraction data, within each of the plurality of time periods, to determine occurrence of a marker selected from the markers consisting of Fetal Evaluation (FE), Fetal Intervention (FI), Fetal Operation (FO), and Fetal Morbidity (FM); and provide a recommended intervention based on the occurrence of the marker.

20. The system according to claim 19, wherein the plurality of time periods includes: a first time period defined as predefined time before the contraction begins to the onset of the contraction, a second time period defined as duration of the contraction, a third time period defined as a predefined period after the contraction ends, and a fourth time period defined as any additional time after the end of the third time period till a predetermined time before the next contraction;21. The system according to claim 19, wherein the recommended intervention is selected from one or more of: no intervention required, improving tracing of the FHR pattern and / or contraction pattern by adjusting or switching the measuring equipment, desisting from pushing and / or desisting from inducing labor, reducing uterine activity by reducing Pitocin and / or administering uterine muscle relaxant, and delivery.

22. The system according to claim 19, wherein the recommended intervention for Fetal Evaluation (FE) includes: examine cervix to determine cervical dilation, effacement, station, fetal headposition, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis, perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE), perform artificial rupture of membranes (AROM), reducing or stopping administration of picotin; place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion; if in 2nd stage (pushing), allow pushing only every other contraction, but if nonengaged do not allow push with each contraction, perform abdominal ultrasound to determine present fetal presentation, begin discussion of informed consent for operative vaginal delivery and begin consent process, inform labor staff of the potential need to perform operative intervention to help assemble the 2nd OR team, and / or evaluate for chorioamnionitis if fetal tachycardia or s / s chori oamnioitis with administration of antibiotics if IAI present or suspected.

23. The system according to claim 19, wherein the recommended intervention for Fetal Intervention (FI) includes: examine cervix to determine cervical dilation, effacement, station, fetal head position, presence of absence of rotation and descent, molding, caput and / or adequacy of pelvis, perform fetal scalp stimulation, including placing a fetal Scalp Electrode (FSE), perform artificial rupture of membranes (AROM) if not already done, cease all pitocin induction / augmentation if being done and pushing efforts if in the 2nd stage of labor, place intrauterine pressure catheter for UC evaluation till delivery and begin manioinfusion, administer 25 cc sc terbutaline if contractions > 6 / 10 minutes or hypertonic uterus, prepare staff for the potential of operative delivery, discuss and signing of operative delivery informed consent,full intrauterine resuscitation (IR) with cessation of all induction agents including pitocin, cervadil and / or cytotec, maternal repositioning, infusion of IV fluids, provide oxygen by mask, amnioinfusion if variable FHR decelerations have not started, move patient to operating room if FI persists more than 10 minutes, and / or prepare and examine patient for possible operative dl by anesthesiologist.

24. The system according to claim 19, wherein the recommended intervention for Fetal Operation (FO) includes: prep patient for operative, induce anesthesia if FHR tracing has not improved within 10 minutes, perform operative delivery, fetal head cooling after delivery, continue FHR monitoring following delivery for 1 hour to observe for characteristics of fetal tachycardia and absent variability (Category III tracing), perform c / section, and / or cord gases and placental pathology.

25. The system according to claim 19, wherein the recommended intervention for Fetal Morbidity includes: full neonatal resuscitation with Fetal scalp electrode on the fetus after delivery, emergency delivery, brain cooling, and / or cord gases and placental pathology.

26. The system according to claim 19, wherein analyzing, the processor is configured to determine fetal features including baseline fetal heart rate (FHR), baseline variability, accelerations, decelerations, fetal recovery, suspected fetal acidosis, and / or fetal behavior.

27. The system according to Claim 19, wherein analyzing, the processor is configured to determine uterine contraction parameters including baseline tone, frequency, amplitude, duration, intervals, rest-time, not interpretable, absent, rhythm, pushing, and / or uterine activity.

28. The system according to claim 19, wherein analyzing, the processor is configured to determine FHR accelerations that increase >10 bpm above the precontraction FHR baseline for at least 15 seconds, and determining FHR decelerations that have FHR values less than the precontraction FHR rate during the contractions and / or after the contraction ends.

29. The system according to claim 19, wherein analyzing, the processor is configured to classify FHR decelerations as early, variable, or late depending on the depth of the deceleration, its timing of onset and return to baseline FHR, symmetry, and repetitiveness.

30. The system according to claim 19, wherein analyzing, the processor is configured to determine the rates of change of deacclerations and / or accelerations.

31. The system according to claim 30, wherein the rates of change include velocity and / or acceleration.

32. The system according to claim 19, wherein analyzing, the processor includes utilization of a neural network.

33. The system according to claim 32, wherein the neural network is trained using predetermined fetal heart rate (FHR) and uterine activity data along with known occurrence of markers.

34. The system according to claim 32, wherein the training is suspended after a predetermined level of accuracy and / or precision is achieved.

35. A method for automated feature extraction, comprising:acquiring a set of one or more non-digitized charts having one or more predetermined features; digitally assigning one or more markers to the one or more predetermined features in the charts; digitally supplying the assigned marker-feature sets to a supervised model; having the model statistically iterate over the assigned marker-feature sets; automatically assessing model performance one or more times during and / or after the statistical iteration process until a predetermined level of the model performance is achieved; and applying the model to one or more new sets of one or more charts to extract one or more non-assigned predetermined features.

36. The method according to claim 35, wherein the set of one or more non-digitized charts includes a time-series set of one or more charts.

37. The method according to claim 35, wherein applying the model to one or more new sets of one or more charts includes applying the model to one or more digitized charts, signal data, and / or tabular data.

38. The method according to claim 35, wherein the supervised model includes a convolution neural network model.

39. The method according to claim 38, wherein the convolution neural network model is configured to implement a basic single-shot detection deep learning approach.

40. The method according to claim 35, wherein the markers include one or more digital pixel coordinates corresponding to a location of the one or more features in the non-digitized charts.

41. The method according to claim 35, wherein the charts comprise one or more numeric arrays of pixel values with varying signal counts associated with the pixel values.

42. The method according to claim 35 further including splitting the assigned feature marker set in at least one training subset and at least one testing subset.

43. The method according to claim 42, wherein the testing subset comprises about 20% of the original assigned marker-feature set.

44. The method according to claim 43, wherein automatically assessing model performance includes observing model stability and / or minimization in its statistical loss function.

45. The method according to claim 43, wherein the model is configured to be integrated in a software application.

46. A method for automated chart processing, comprising: analyzing one or more time-series sets of one or more non-digitized charts of fetal heart rate (FHR) and / or concurrent maternal uterine contractions (UC) to digitally associate one or more markers with one or more predetermined fetal signatures in the charts; automatically transmitting the associated marker-signature groups to an artificial intelligence model; using the transmitted groups to perform supervised training of the artificial intelligence model; determining accuracy and / or precision of the model by comparing one or more predictions produced by the model to one or more previously unfamiliar to the model time- series sets of one or more charts of FHR and / or concurrent maternal UC with associated marker-signature groups; suspending the model training after a predetermined level of accuracy and / or precision is achieved; and applying the trained model to automatically process one or more new time-series sets of one or more charts of FHR and / or concurrent maternal UC, having one or more un- associated predetermined fetal signatures.

47. The method according to claim 46, wherein the artificial intelligence model includes a convolution neural network model configured to implement a basic single-shot detectiondeep learning approach.

48. The method according to claim 46, wherein the supervised training of the artificial intelligence model includes statistically iterating over the transmitted groups to establish one or more associations between the markers and / or marker patterns in the FHR and / or UC with the one or more predetermined fetal signatures.

49. The method according to claim 46, wherein the time-series sets are obtained from non- digitized charts.

50. The method according to claim 46, further including additional model training to include user feedback and / or provide additional training data.

51. The method according to claim 46, wherein the fetal features include fetal heart rate parameters including baseline fetal heart rate (stable, rising, falling, bradycardia, tachycardia, not defined, and / or uninterpretable), baseline variability (moderate, increased, decreased, and / or absent), accelerations (present, absent), decelerations (absent, early, late, prolonged, not identified), fetal recovery (normal, abnormal / overshoot, abnormal / delayed, abnormal / tachycardia), suspected fetal acidosis, and / or fetal behavior (normal - term, normal - preterm (later), abnormal, abnormal - suspected fetal injury).

52. The method according to claim 46, wherein the fetal features include uterine contraction parameters including baseline tone (normal range, increased / increasing, decreased / decreasing, not interpretable, and / or absent), frequency, amplitude, duration, intervals (between contractions, peak to peak), rest-time, not interpretable, absent, rhythm (no pattern, regular, irregular), pushing (absent, present), uterine activity (absent, uninterpretable, excessive, tachysystole, hypertonus, coupling / tripling, and / or rest time - decreased).

53. The method according to claim 46, further including additional processing of the one or more fetal signatures processed with the trained model.

54. The method according to claim 53, further including generating a report and / or producing a notification for a user including an intervention recommendation (no intervention required, improve tracing of the FHR pattern (adjust the equipment, consider internal electrode, switch to an external device) and / or contraction pattern (adjust the equipment, consider an internal device), conservative actions (desist from pushing, desist from inducing labor, reduce uterine activity (reduce Pitocin, administer uterine muscle relaxant), consider delivery, adjust measuring equipment).

Images