Search

CN-121987246-A - System for generating physiological time series via deep learning model detection or prediction

CN121987246ACN 121987246 ACN121987246 ACN 121987246ACN-121987246-A

Abstract

Methods and systems are provided for detecting and predicting Uterine Activity (UA) from ultrasound receive signals and/or Electrocardiogram (EHG) data using a neural network model (420) trained using ultrasound time series data (402) and/or EHG data as input data (404) and time series tocodynamometer (toco) data as truth data (405). After training, the trained UA detection model (430) may be used to detect the timing, duration, and intensity of uterine contractions in a new patient with greater accuracy than toco. The output of the trained UA detection model (430) may also be used to train a UA prediction model (470) to predict the next uterine contraction (1210) based on a series of previous uterine contractions. Thus, uterine contractions may be detected and/or predicted during Electronic Fetal Monitoring (EFM) without the use of additional toco devices, thereby reducing the cost of EFM and the number of sensing devices upon which EFM depends.

Inventors

  • M. Kyle
  • K. Manican

Assignees

  • 通用电气精准医疗有限责任公司

Dates

Publication Date
20260508
Application Date
20251024
Priority Date
20241107

Claims (15)

  1. 1. A method, the method comprising: acquiring ultrasound time series data of the abdomen and/or fetus of the pregnant mother (602); Inputting the ultrasound time series data into a first trained neural network model (606); generating a plot (610) of Uterine Activity (UA) of the mother or the fetus from the output of the first trained neural network model, and The drawing (610) is displayed on a display device.
  2. 2. The method of claim 1, wherein the ultrasound time series data is acquired from a received signal of an ultrasound probe prior to processing the ultrasound time series data to generate an ultrasound physiological signal.
  3. 3. The method of claim 2, wherein the ultrasound physiological signal is displayed on the display device simultaneously with the mapping.
  4. 4. The method of claim 1, wherein the first trained neural network model is trained by using stored ultrasound time series data acquired from a plurality of subjects as input data and using stored time series data of the maternal abdomen acquired from the same plurality of subjects at the same time as the stored ultrasound time series data using a tocodynamometer as truth data.
  5. 5. The method of claim 1, wherein the first trained neural network model is trained by using stored time series data of the maternal abdomen and/or fetus acquired from a plurality of subjects using an Electrocardiogram (EHG) device as input data and using stored time series data of the maternal abdomen and/or fetus acquired from the same plurality of subjects at the same time as the stored EHG time series data using a tocodynamometer as truth data.
  6. 6. The method of claim 1, wherein inputting the ultrasound time series data into the first trained neural network model further comprises: Generating a plurality of overlapping ultrasound time series windows (604) of the ultrasound time series data in predefined time increments, and An overlapping ultrasound time series window of the plurality of overlapping ultrasound time series windows is input into the first trained neural network model at each time increment (606).
  7. 7. The method of claim 6, wherein generating a plot of the UA of the mother from the output of the first trained neural network model further comprises: Receiving a plurality of overlapping windows of UA time series data as output of the first trained neural network model, each overlapping window of UA time series data being generated in the predefined time increment; Aggregating the plurality of overlapping windows of the UA time series data by averaging values associated with the same time in each overlapping window to create a continuous stream of UA time series data (608); the plot of the UA is generated from a continuous stream of the UA time series data (610).
  8. 8. The method of claim 7, the method further comprising: Generating training data for a second trained neural network model from the plurality of overlapping windows of UA time series data, the training data comprising a training pair comprising an overlapping window of UA time series data as input data and a portion of a subsequent overlapping window of UA time series data as truth data (508), and The training data is used to train the second trained neural network to predict a subsequent uterine contraction from a series of previous uterine contractions (512).
  9. 9. The method of claim 1, wherein the first trained neural network model comprises one of a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN).
  10. 10. The method of claim 9, wherein the first trained neural network model is a CNN and the first layer of the CNN is a wavelet decomposition transform layer that decomposes an input signal of the CNN into dominant frequency components.
  11. 11. A Uterine Activity (UA) detection system (302), the UA detection system comprising: display device (334) A processor (304) communicatively coupled to the display device (334) and a non-transitory memory (306) comprising instructions that, when executed, cause the processor (304) to: During a first training phase of a UA detection model (420) of the UA detection system (302): Generating a first plurality of overlapping windows (408) of ultrasound time series data (402) acquired from a plurality of ultrasound scans performed on a plurality of fetuses and/or abdomen of a pregnant mother; Generating a second plurality of overlapping windows (408) of UA time series data (402) acquired from the pregnant mother via a tocodynamometer (toco) at the same time as the ultrasound time series data; Training the UA detection model (420) on a training pair (412) comprising a time-series window (408) of the first plurality of overlapping windows as input data (404) and a time-series window of the second plurality of overlapping windows as truth data (405), and During a second inference phase of the UA detection model (420): receiving ultrasound time series data (456) from an ultrasound probe placed over an abdomen (452) of a fetus or pregnant mother; generating a third plurality of overlapping windows (458) of the ultrasound time series data; Inputting overlapping windows (458) of the third plurality of overlapping windows (458) into the trained UA detection model (430) at predefined time increments; Receiving a fourth plurality of overlapping windows of UA time series data (460) as output of the trained UA detection model (430) at each predefined time increment; Aggregating the fourth plurality of overlapping windows of UA time series data (460) by averaging values associated with the same time in each overlapping window to create a continuous stream of UA time series data (464), the continuous stream of UA time series data (464) showing uterine contractions of the pregnant mother (452), and Displaying a plot of the continuous stream of UA time series data on the display device.
  12. 12. The UA detection system (302) of claim 11, wherein the ultrasound time series data (456) is acquired from a received signal of an ultrasound probe prior to processing the ultrasound time series data (456) to generate an ultrasound physiological signal.
  13. 13. The UA detection system (302) of claim 11, wherein further instructions are stored in the memory (306) that, when executed, cause the processor (304) to process the ultrasound time series data (456) to generate an ultrasound physiological signal (720) and display the ultrasound physiological signal (720) on the display device (334) concurrently with the drawing (722) of the continuous stream (464) of UA time series data.
  14. 14. The UA detection system (302) of claim 11, wherein the UA detection model (420) is trained using UA time-series data (402) acquired from the pregnant mother via an Electrocardiogram (EHG) device during the plurality of ultrasound scans as truth data.
  15. 15. The UA detection system (302) of claim 11, wherein further instructions are stored in the memory (306), the further instructions when executed causing the processor (304) to: generating training data for a UA prediction model (420) from the plurality of overlapping windows (408) of UA time-series data (402), the training data comprising training pairs (412) comprising overlapping windows (408) of UA time-series data (402) as input data and a portion of subsequent overlapping windows (408) of UA time-series data (402) as truth data (405), and The UA predictive model (420) is trained using the training data to predict a subsequent uterine contraction (1210) from a series of previous uterine contractions (1202, 1204, 1206, 1208).

Description

System for generating physiological time series via deep learning model detection or prediction Technical Field Embodiments of the subject matter disclosed herein relate to detecting or predicting physiological signals of a patient, including fetal heart rate activity and uterine activity in a pregnant mother. Background Uterine Activity (UA) corresponding to the uterine contractions of the mother may be measured using a tocodynamometer or toco, which measures displacement of the mother's abdomen via a mechanical pressure transducer mounted on the belt toco. UA may be represented as time series data that may be used to determine the frequency, amplitude, duration, and/or period of uterine contractions. However, due to the dependence on abdominal position and variable tension on the belt and the altitude ambient pressure dynamics that may affect the data, the data generated by toco may not be as accurate as desired. Clinical ultrasound is an imaging modality that uses ultrasound waves to detect internal structures of a patient's body and generate corresponding physiological signals. For example, an ultrasound probe comprising a plurality of transducer elements emits ultrasound pulses that are reflected or returned, refracted or absorbed by structures in the body. The ultrasound probe then receives reflected echoes, which are processed into signals. Transabdominal ultrasound performed on the pregnant mother also captures the contraction of uterine muscles positioned in the path of the abdominal ultrasound beam. However, the time series data generated by toco may not be as accurate as trans-abdominal ultrasound to indicate the frequency and duration of uterine contractions. Ultrasound has more sensitivity to tissue compression and less positional dependence than Toco when the transducer is placed on the mother's abdomen. Disclosure of Invention The present disclosure solves one or more of the above-identified problems, at least in part, via a method comprising acquiring ultrasound time series data of an abdomen and/or fetus of a pregnant mother, inputting the ultrasound time series data into a first trained neural network model, generating a plot of maternal Uterine Activity (UA) from an output of the first trained neural network model, and displaying the plot on a display device. The above advantages and other advantages and features of the present description will be apparent from the following detailed description when taken alone or in conjunction with the accompanying drawings. It should be understood that the above summary is provided to introduce in simplified form a set of concepts that are further described in the detailed description. This is not intended to identify key features or essential features of the claimed subject matter, the scope of which is defined uniquely by the claims that follow the detailed description. Furthermore, the claimed subject matter is not limited to implementations that solve any disadvantages noted above or in any part of this disclosure. Drawings Various aspects of the disclosure may be better understood by reading the following detailed description and by reference to the drawings in which: FIG. 1 illustrates a block diagram of an exemplary ultrasound system; FIG. 2 illustrates an exemplary tocodynamometer; FIG. 3 is a schematic diagram of an exemplary UA detection system for predicting uterine contractions in a pregnant mother; fig. 4 is an information flow diagram illustrating an exemplary system for training and deploying a UA detection model for use by a UA detection system; fig. 5 is a flow chart illustrating an exemplary method for training a UA detection model; fig. 6 is a flowchart illustrating a method for predicting uterine contractions using a UA detection model; Fig. 7 is a first graph showing time series data of UA in a pregnant mother as prior art; fig. 8 is a second graph showing time series data of UA in a pregnant mother as prior art; fig. 9 illustrates an exemplary array of UA data values extracted from UA time series data; Fig. 10 shows an example of a UA time sequence window extracted from an array of UA data values; Fig. 11 is a third graph showing time series data of UA in a pregnant mother as prior art; FIG. 12 is a graph illustrating a plot of an exemplary output of the UA detection model, an Fig. 13 is a schematic diagram illustrating an exemplary neural network architecture of a UA detection model. Detailed Description Accurately assessing the health of a fetus during parturient and delivery ensures the health of the neonate and mother. Electronic Fetal Monitoring (EFM) is a common obstetric procedure in the united states, and most women receive EFM during pregnancy, parturient and childbirth. The rationale for the use of EFM assumes that Fetal Heart Rate (FHR) abnormalities accurately reflect the risk of hypoxia (hypoxia in the fetus) and that early identification of hypoxia may prompt intervention to improve the outcome of the fetus. EFM can