Neural Network Desaturation Severity Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biomedical monitoring devices, such as pulse oximeters, provide lagging indicators of physiological conditions, making it challenging to promptly identify the severity of oxygen level desaturation events.
Innovation Solution
A method using a neural network to classify oxygen level desaturation events as severe or non-severe by analyzing input signal sequences and feature matrices derived from pulse oximeter data, thereby adjusting alarm delays accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional alarm systems are used to monitor oxygen saturation levels, then all desaturation events are detected, but alarm delays occur for non-severe events and clinical resources are wasted on false alarms
Solution Approach 1:
The system performs preliminary classification of desaturation events using a neural network model that analyzes multiple features (depth, duration, slope) before triggering an alarm. This preliminary action distinguishes between severe and non-severe events, allowing the system to apply different alarm strategies in advance, thereby reducing unnecessary alarm delays for non-severe events while maintaining rapid response for severe events.
Solution Approach 2:
The alarm delay is made dynamic rather than fixed. The system adjusts the alarm delay based on the predicted severity of the desaturation event. For severe events, the alarm is triggered immediately with minimal delay, while for non-severe events, a longer delay is applied. This dynamic adjustment resolves the contradiction by adapting the response time to the actual clinical need.
2Object-generated harmful factors
If alarm delay is increased to filter out non-severe events, then false alarms are reduced, but response time to severe events may be delayed
Solution Approach 1:
The system performs preliminary classification using a neural network that evaluates multiple features (depth, duration, slope) before determining alarm timing. This preliminary analysis identifies severe events early, allowing the system to prepare for immediate alarm activation without waiting for the full desaturation event to complete, thus maintaining fast response while filtering non-severe events.
Solution Approach 2:
The alarm delay is dynamically adjusted based on predicted severity. For events classified as severe, the alarm delay is minimized or eliminated entirely, ensuring rapid response. For non-severe events, a longer delay or no alarm is applied. This dynamic strategy reduces false alarms without compromising response speed to critical events.
3Productivity
If neural network classification is implemented to predict desaturation severity, then alarm timing is optimized, but computational complexity and processing time increase
Solution Approach 1:
The neural network model is segmented into a set of predefined rules based on depth, duration, and slope thresholds. Instead of using a complex black-box neural network, the system divides the classification task into discrete feature evaluations, each with its own threshold. This segmentation reduces computational complexity while maintaining the ability to distinguish severe from non-severe events.
Solution Approach 2:
The system changes the parameters used for classification from raw signal data to derived features (depth, duration, slope). By preprocessing the signal to extract these key parameters before classification, the system reduces the dimensionality and complexity of the input data for the neural network, making processing more efficient while retaining the essential information needed for severity prediction.
Data Source
AI summary
Implementations described herein disclose a method of classifying oxygen level desaturation events. In one implementation, the method includes receiving input signal sequences, the input signals indicative of a physiological condition of a patient, generating an input sequence of oxygen saturation levels based on the input signal sequence, comparing the input sequence of oxygen saturation levels to a desaturation alarm threshold to determine a desaturation event, generating an input feature matrix based on at least one of the input signal sequences and the input sequence of oxygen saturation levels, and classifying based on the input feature matrix, using a neural network, the desaturation event being a severe desaturation event (SDE) or a non-severe desaturation event (non-SDE).