PET Respiratory Gating Using PCA to Remove Cardiac Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing PET imaging systems face challenges in accurately separating cardiac and respiratory signals during data acquisition, leading to image blurring and artifacts due to the inclusion of cardiac signals in respiratory motion correction, which is typically addressed using external motion sensors that can be cumbersome and synchronization issues.
Innovation Solution
A data-driven approach that involves dividing PET data into short frames, determining cardiac cycle length, re-binning with overlap to include full cardiac cycles, applying PCA to separate cardiac and respiratory waveforms, and reconstructing images using the separated waveforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If external motion sensors are attached to detect biosignals for gating, then respiratory motion can be detected, but the system becomes more cumbersome and synchronization issues may occur
Solution Approach 1:
The PET imaging system extracts respiratory and cardiac signals directly from its own acquired PET data without requiring external sensors. The system serves itself by utilizing the temporal variations in coincidence events during list-mode acquisition to derive biosignals through PCA, eliminating the need for separate motion detection devices and their associated synchronization complexities
Solution Approach 2:
The respiratory and cardiac signals are extracted directly from the PET coincidence event data using principal component analysis. By analyzing temporal variations in the acquired PET data across multiple frames, the system separates and extracts the underlying physiological signals without requiring external sensing equipment
2Measurement precision
If PCA is applied on short frames to extract biosignals, then respiratory motion can be captured, but cardiac signals are also included causing signal contamination
Solution Approach 1:
The extracted biosignal is segmented into respiratory and cardiac components by analyzing frequency characteristics. After initial PCA extraction, the signal is divided into frequency bands where lower frequencies correspond to respiratory motion and higher frequencies correspond to cardiac beating, allowing separate reconstruction for each physiological process
Solution Approach 2:
The method utilizes the periodic nature of respiratory and cardiac cycles to separate signals. By analyzing the temporal periodicity and frequency content of the extracted biosignal, the system identifies and separates the slower periodic respiratory waveform from the faster periodic cardiac waveform through frequency-based filtering
3Ease of manufacture
If PET data is divided into short non-overlapping frames, then processing is simplified, but cardiac cycles may be fragmented across frames
Solution Approach 1:
The system performs preliminary determination of cardiac cycle length by analyzing initial short frames, then uses this information to configure subsequent overlapping frame structures. This preliminary analysis allows the system to set appropriate frame parameters that ensure complete cardiac cycles are captured within frames while maintaining processing efficiency
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 method effectively separates cardiac and respiratory signals, improving image quality by reducing motion blur and artifacts, achieving results comparable to systems using external devices for respiratory measurement.
Implementation Method 1
applying a principal component analysis (PCA) process on the re-binned list mode data having the second frame length to determine a respiratory waveform
Data Source
AI summary
A method for signal separation includes obtaining list mode data representing radiation detected during an imaging scan, the list mode data being affected by quasi-periodic motion of an imaging object; dividing the list mode data into first non-overlapping frames of a first frame length, and process the first frames to determine a cardiac cycle length; determining a second frame length, longer than the first frame length, based on the determined cardiac cycle length; re-binning the list mode data into overlapping frames having the second frame length, based on the non-overlapping frames having the first frame length; applying a principal component analysis (PCA) process on the re-binned list mode data having the second frame length to determine a respiratory waveform; determining a cardiac waveform using the determined respiratory waveform; and reconstructing an image based on the list mode data using the determined respiratory waveform and the determined cardiac waveform.