ECG Cardiac Event Classification Using ML for Normal Rhythm Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The overwhelming number and frequency of potential cardiac events detected by medical devices can burden physicians with unnecessary data processing, including instances of normal cardiac activity, which can lead to resource inefficiencies.
Innovation Solution
A cardiac event evaluation system using a classification model, such as an ensemble of boosted trees, processes ECG waveforms to assess potential cardiac events, determining normal rhythms and generating alerts on mobile devices, thereby intercepting and preventing unnecessary physician notifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all potential cardiac events are forwarded to physicians for review, then no diagnostic events are missed, but physician workload increases significantly and resources are wasted on normal rhythms
Solution Approach 1:
A machine learning classification model serves as an intermediary between the medical device and the physician. The device first captures ECG waveforms, then the ML model classifies them as normal or abnormal before forwarding to the physician. This intermediary layer filters out normal rhythms, allowing physicians to focus only on abnormal cases while maintaining diagnostic accuracy.
Solution Approach 2:
The system performs preliminary classification of ECG waveforms using a trained machine learning model before physician review. By pre-processing and categorizing events at the device level, the system prepares data in advance, separating normal from abnormal rhythms so physicians receive only relevant cases requiring their expertise.
2Loss of information
If ECG waveforms are processed and transmitted for every detected event, then complete data is available for review, but data transmission and storage requirements increase
Solution Approach 1:
The system extracts and transmits only essential features from ECG waveforms rather than complete raw data. The machine learning model processes waveforms locally, extracting classification results and key interval data (such as R-R intervals) for transmission. This selective extraction maintains diagnostic utility while significantly reducing data transmission and storage requirements.
3Productivity
If a simple filtering approach is used to reduce events, then processing is faster, but false positives increase and normal events may be missed
Solution Approach 1:
The system replaces simple mechanical filtering rules with an intelligent machine learning classification model. Instead of using fixed thresholds or basic signal processing filters, the ML model learns complex patterns from training data, enabling accurate distinction between normal and abnormal rhythms while maintaining high processing speed. This substitution of mechanical filtering with intelligent classification resolves the trade-off between speed and accuracy.
Data Source
AI summary
Systems, methods, and devices involve receiving a package of data that includes an electrocardiogram (ECG) waveform associated with a potential cardiac event, processing the ECG waveform to extract interval data associated with the potential cardiac event, inputting the ECG waveform and the interval data into a trained machine learning model, and determining, by the trained machine learning model, that the potential cardiac event comprises a normal rhythm.