PET Image Normalization via Peak-Alignment to Remove Reference Bias
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Standardized Uptake Value Ratio (SUVR) methods for PET scans are biased due to variations in reference region signals, particularly in tau PET studies, leading to underestimation or overestimation of real group differences and limitations in distinguishing positive uptake, which affects the interpretation of disease progression and treatment efficacy.
Innovation Solution
The method involves localizing PET data using anatomical masks, generating normalized image intensity values through peak-alignment normalization, which standardizes SUVs by subtracting the peak value from each SUV value divided by the spread, thereby reducing bias and providing a robust analysis of regional distribution patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional SUVR normalization method is used, then PET scan data can be normalized using a reference region, but variations in reference region signal cause bias leading to underestimation or overestimation of real group differences
Solution Approach 1:
The patent extracts the reference region signal from the normalization process entirely. Instead of using cerebellar gray matter as a reference, the method uses each voxel's own signal intensity distribution through histogram analysis and Gaussian fitting. This removes the source of bias (reference region variations) while maintaining normalization capability through the formula N=(P-M)/S where P is pixel intensity, M is Gaussian mean, and S is Gaussian standard deviation.
Solution Approach 2:
The patent introduces a Gaussian distribution model as an intermediary between raw PET signal and normalized values. By fitting a Gaussian curve to the histogram of pixel intensities and using its parameters (mean M and standard deviation S) for normalization, the method mediates the transformation in a way that eliminates reference region dependency while preserving relative signal differences.
2Adaptability or versatility
If SUVR method is used, then PET data can be normalized, but the range and cutoff values vary among different tracers and PET scanners making it challenging to distinguish positive uptake
Solution Approach 1:
The patent changes the normalization parameters from fixed reference region values to dynamic parameters derived from each image's own signal distribution (Gaussian mean M and standard deviation S). This adaptation allows the method to work across different tracers and scanners because each image is normalized relative to its own characteristics rather than an external reference, enabling consistent positive uptake detection universally.
3Ease of operation
If conventional reference region normalization is used, then PET scan analysis can be performed, but off-target binding in reference region affects the validity of the normalization
Solution Approach 1:
The patent removes the reference region from the analysis entirely, extracting only the relevant signal information from each voxel through histogram analysis. By eliminating the cerebellar reference region that suffers from off-target binding, the method maintains analytical simplicity while ensuring normalization validity is not compromised by invalid reference signals.
Data Source
AI summary
Systems and methods are for analyzing Positron Emission Tomography (PET) image data. The methods may include generating a set of standardized uptake values (SUVs) of global or localized PET data for voxels within a selected region of interest (ROI), normalizing the set of SUVs by generating a set of SUVPs where each corresponding SUVP for each SUV is obtained using the formula: SUVP=(SUV−M)/S, wherein M corresponds to a peak value for the set of SUVs, and S corresponds to a spread for the set of SUVs, and generating a normalized image based on the set of SUVPs for the ROI. The systems may include any suitable device for PET image analysis performing the methods.