PET Image Gating Using Latent Motion Features From Short Segments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing PET imaging systems suffer from image blurring and artifacts due to patient motion, particularly from involuntary movements like respiration and heartbeat, leading to inaccurate lesion volume estimation and activity underestimation, and external motion sensors complicate the scanning process.
Innovation Solution
A data-driven gating method using an autoencoder network to extract latent features from sinogram image data, segmenting the data into shorter intervals, and clustering these features to reconstruct images based on similar motion phases, reducing the need for external motion trackers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If external motion sensors are attached to patients during PET scan, then motion information can be recorded for gating, but the scanning process becomes more cumbersome and complex
Solution Approach 1:
The system uses the PET scan data itself to extract motion information through automated algorithms (ICA or PCA), eliminating the need for external motion sensors. The data serves dual purposes: both imaging and motion detection, making the system self-sufficient and reducing external dependencies.
Solution Approach 2:
The patent extracts motion information directly from the PET scan data by separating motion-related components from the imaging data using signal separation techniques. This extraction process obtains necessary motion information without adding external sensing equipment.
2Device complexity
If signal separation techniques such as ICA or PCA are applied for data-driven gating, then motion information can be extracted without external sensors, but the processing time increases
Solution Approach 1:
The patent divides the PET scan data into multiple time segments or frames, applying signal separation techniques to each segment individually. This segmentation allows for more manageable processing while maintaining the ability to detect motion across the entire scan duration.
Solution Approach 2:
The system performs preliminary processing by pre-segmenting the data and pre-identifying potential motion components before final reconstruction. This preliminary action prepares the data in advance, reducing the computational burden during the final gating determination.
3Stability of the object's composition
If image data is segmented into shorter durations, then motion within each segment is reduced, but the number of segments increases requiring more processing
Solution Approach 1:
The patent merges adjacent image segments that exhibit similar motion characteristics by clustering analysis. This combining process reduces the total number of segments that need to be processed individually while maintaining motion consistency within each merged group, thereby improving processing efficiency.
Solution Approach 2:
The system uses feedback from motion analysis of one segment to inform the processing of subsequent segments. By learning from previously analyzed segments, the system can make informed decisions about segment merging and gating strategies, reducing redundant processing.
Data Source
AI summary
A method, apparatus, and computer instructions stored on a computer-readable medium perform latent image feature extraction by performing the functions of receiving image data acquired during an imaging of a patient, wherein the image data includes motion by the patient during the imaging; segmenting the image data to include M image data segments corresponding to at least N motion phases having shorter durations than a duration of the motion by the patient during the imaging, wherein M is a positive integer greater than or equal to a positive integer N; producing, from the M image data segments, at least N latent feature vectors corresponding to the motion by the patient during the imaging; and performing a gated reconstruction of the N motion phases by reconstructing the image data based on the at least N latent feature vectors.