ST Segment Detection in ECG Signals Using AI Baseline Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the ST segment of an electrocardiogram signal are inaccurate when there is significant interference or drift, leading to reduced accuracy in identifying myocardial ischemia and infarction.
Innovation Solution
An AI-based method that uses wavelet transformation and triangular area methods to locate key features, followed by average filtering to establish a standard baseline, allowing for precise identification of the ST segment and determination of its elevation, depression, or normal status.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If time window detection method or wavelet filtering method is used to extract and identify ST segment, then calculation is simple and implementation is easy, but accuracy in identification and determination of ST segment is greatly reduced when there is significant interference or drift effects
Solution Approach 1:
The patent applies preliminary action by first performing wavelet transformation to obtain a wavelet coefficient sequence, then performing partial enlargement on specific segments (QRS complex, T wave) before identifying feature points. This preprocessing sequence prepares the signal in advance to enhance feature visibility and improve subsequent ST segment determination accuracy even in noisy conditions
Solution Approach 2:
The patent replaces traditional mechanical/time-window-based detection methods with a signal processing approach using wavelet transformation. This substitution transforms the problem from fixed-window detection to frequency-time domain analysis, enabling accurate feature point identification despite signal interference or drift
2Device complexity
If traditional methods are used for ST segment detection, then computational complexity is low, but reliability is reduced under noisy conditions
Solution Approach 1:
The patent segments the electrocardiogram signal into distinct components (QRS complex, T wave, ST segment) and applies different processing strategies to each. By using partial enlargement on specific segments and identifying feature points separately, the method improves reliability of ST segment detection while keeping computational complexity manageable through focused processing
3Measurement precision
If feature point identification accuracy is improved through multiple processing steps, then ST segment determination accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies partial action by performing wavelet transformation and partial enlargement only on specific segments of the signal (QRS complex, T wave) rather than the entire signal. This selective processing maintains high accuracy in feature point identification while reducing overall processing time compared to full-signal processing
Data Source
Figure 1A~1C
Figure 2
Figure 3
AI summary
The present invention provides an AI-based method and device for automatically determining an ST segment of an electrocardiogram signal. With respect to the filtered human electrocardiogram signal, firstly, the key feature points of the electrocardiogram signal are separately extracted by combining the wavelet filtering and the triangular area method, including: locating the S wave and the T wave, and accurately identifying the start point and the slope characteristic of the ST segment. Secondly, a method for removing the baseline based on the average filtering is provided to subtract the extracted baseline sequence to obtain a new electrocardiogram signal. With respect to the new electrocardiogram signal, the ST segment and each baseline segment are extracted, and the heart rate and the slope of each baseline segment are calculated, and a standard baseline is comprehensively determined. Finally, the abnormal changes of the ST segment are identified based on the standard baseline to obtain the qualitative and quantitative determination result of the ST segment whether it is elevated, depressed, or normal. The present invention solves the problems of the baseline error caused by the baseline drift, the inaccurate selection of the standard baseline, and the large difference in the number of abnormal heart beats, thereby indirectly improving the accuracy of the determination of the abnormal changes of the ST segment. The present invention has a simple calculation and can be easily implemented.