Related Experiment Video
Updated: Jun 25, 2026
06:33
Construction of a Preclinical Multimodality Phantom Using Tissue-mimicking Materials for Quality Assurance in Tumor Size Measurement
Published on: July 29, 2013
Kernel-based maximum likelihood reconstruction of attenuation and activity (MLAA) in SPECT imaging for improved
Chenguang Li1,2, Yansong Zhu3, Lucas Alexander Polson1,2
1Department of Physics & Astronomy, The University of British Columbia, Vancouver, Canada.
Physics in Medicine and Biology
|June 23, 2026
Summary
A new kernel-based maximum-likelihood attenuation and activity (MLAA) algorithm improves quantitative SPECT/CT imaging by enhancing attenuation map accuracy. This kernel MLAA method significantly reduces bias and artifacts, offering more reliable radiopharmaceutical therapy quantification.
Area of Science:
- Medical Imaging
- Nuclear Medicine
- Computational Imaging
Background:
- Quantitative SPECT/CT for radiopharmaceutical therapy requires accurate attenuation maps (µ-maps).
- CT-derived µ-maps require energy-dependent conversion, often using bilinear scaling (BLS), which can introduce quantification bias.
- Existing methods struggle with attenuation-activity crosstalk, impacting accuracy.
Purpose of the Study:
- To develop and evaluate a kernel-based maximum-likelihood attenuation and activity (MLAA) algorithm for improved µ-map accuracy in quantitative SPECT/CT.
- To address the limitations of standard MLAA and CT-based attenuation correction methods.
- To enhance the reliability of activity quantification in radiopharmaceutical therapy.
Main Methods:
- Developed a kernel MLAA algorithm that jointly estimates µ-map and activity distribution using Poisson log likelihood maximization.
- Incorporated CT-derived anatomical features into the MLAA algorithm to mitigate attenuation-activity crosstalk.
- Validated the method using Monte Carlo simulations, phantom studies, and patient data, implemented in the GPU-accelerated PyTomography platform.
Main Results:
- Kernel MLAA significantly suppressed crosstalk artifacts and reduced µ-map bias compared to standard MLAA and BLS, decreasing liver voxel-wise MAE by ~84% in simulations.
- Improved activity quantification accuracy, with recovery-coefficient bias converging to 0.90%±0.34% and simulated contrast-induced bias decreasing from 18.11% to 1.51%.
- Restored perturbed µ-values in phantom and patient data, reducing mean bias from 13.6% to 0.6% and 13.7% to 0.3%, respectively, while controlling noise.
Conclusions:
- Kernel MLAA offers a robust solution for improving attenuation correction in quantitative SPECT.
- The algorithm automatically detects and corrects µ-map bias, providing a reliable alternative when CT-derived maps are compromised.
- This advancement has significant implications for accurate activity quantification in radiopharmaceutical therapy.
