Fetal Oximetry Model Training Using Simulated Light Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for monitoring fetal health, such as fetal heart rate monitoring, are inefficient in determining fetal distress and often provide false positive results, leading to unnecessary interventions like Cesarean deliveries, while obtaining sufficient data for training fetal oximetry models is difficult due to the rarity of disease states and invasive data collection methods.
Innovation Solution
A two-step process using simulated light transmission data and machine learning to train a fetal oximetry model, followed by adaptation to an in vivo model, reducing the need for costly and invasive data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fetal heart rate monitoring is used to monitor fetal health, then fetal health monitoring is provided, but the method is inefficient in determining fetal distress and provides false positive results
Solution Approach 1:
The patent uses simulated light transmission data that copies real physiological conditions to train the fetal oximetry model. This simulated data allows the model to learn from numerous virtual examples of fetal distress without requiring equivalent real-world invasive measurements, thereby improving reliability while maintaining monitoring efficiency
Solution Approach 2:
The patent performs preliminary training of the fetal oximetry model using simulated data before deploying it for actual fetal monitoring. This preliminary action with synthetic data prepares the model to accurately distinguish true fetal distress from false positives, improving the reliability of subsequent real-world monitoring without compromising efficiency
2Measurement precision
If in vivo data collection is performed to train fetal oximetry models, then model accuracy is improved, but the process is costly, invasive, and time-consuming
Solution Approach 1:
The patent performs preliminary model training using simulated light transmission data before conducting any in vivo data collection. This preliminary action with synthetic data establishes a baseline model that can then be refined with limited real data, significantly reducing the time and invasive procedures required compared to training solely with in vivo data
Solution Approach 2:
The patent creates copies of real physiological conditions through simulated light transmission data that mimics actual fetal monitoring scenarios. These simulated copies provide abundant training examples without requiring equivalent real-world invasive measurements, improving model accuracy while avoiding the time loss and ethical issues of extensive in vivo data collection
3Measurement precision
If in vivo data collection is performed to train fetal oximetry models, then model accuracy is improved, but the cost and invasiveness increase
Solution Approach 1:
The patent uses simulated light transmission data that copies real fetal monitoring conditions to train the model. This approach provides abundant training data without requiring invasive procedures on actual fetuses, thereby improving model accuracy while eliminating the harmful invasive effects associated with traditional in vivo data collection methods
Solution Approach 2:
The patent introduces simulated data as an intermediary between theoretical model development and actual in vivo testing. This intermediary layer allows the model to be trained on realistic data without direct invasive contact with subjects, reducing harmful factors while maintaining measurement precision
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly shortens the timeline and improves the accuracy of fetal oximetry model development by leveraging simulated data, allowing for more accurate prediction of fetal distress without the need for extensive in vivo data collection.
Implementation Method 1
transmitting an optical signal into the pregnant mammal's abdomen
Implementation Method 2
received a plurality of detected electronic signals that correspond to light emitted from a pregnant mammal's abdomen and a fetus contained therein that has been detected by the detector and converted into the detected electronic signal
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
A plurality of sets of simulated optical inputs that is simulated to travel through an animal model of tissue, thereby generating simulated light transmission data, and corresponding oximetry vales may be used to train a simulated fetal oximetry model to predict, or calculate, oximetry values for subsequently received sets of simulated light transmission data. The simulated fetal oximetry model may be adapted to train an in vivo fetal oximetry model that may be configured to predict, or calculate, fetal oximetry values for subsequently received sets of light transmission data received from an in vivo study of a pregnant mammal and her fetus.