4D numerical observer for lesion detection in respiratory-gated PET
Auranuch Lorsakul1, Quanzheng Li2, Cathryn M Trott3
1Center for Advanced Medical Imaging Sciences, Division of Nuclear Medicine and Molecular Imaging, Massachusetts General Hospital, Boston, Massachusetts 02114 and Department of Biomedical Engineering, Columbia University, New York, New York 10027.
Medical Physics
|October 6, 2014
Summary
A novel 4D numerical observer significantly enhances lesion detection in pulmonary oncology PET scans by integrating spatial and temporal data. This advanced method offers a more reliable assessment of respiratory-gated PET improvements compared to traditional 3D approaches.
Area of Science:
- Medical Imaging Physics
- Radiological Sciences
- Computational Imaging
Background:
- Respiratory-gated positron emission tomography (PET)/computed tomography (CT) protocols improve lesion detection by synchronizing data acquisition, reducing motion artifacts.
- Objective assessment of respiratory-gated PET image quality has been primarily limited to 3D analysis, potentially overlooking temporal information.
- Accurate lesion detection in pulmonary oncology is crucial for effective diagnosis and treatment planning.
Purpose of the Study:
- To propose and evaluate a 4D numerical observer model that incorporates both spatial and temporal information for improved lesion detection in pulmonary oncology.
- To provide a more objective and comprehensive assessment of image quality improvements offered by respiratory-gated PET compared to conventional 3D methods.
Main Methods:
- A 4D numerical observer was developed, combining a 3D channelized Hotelling observer for spatial analysis and a temporal Hotelling observer.
- Simulated (18)F-fluorodeoxyglucose PET data were generated using a 4D anthropomorphic phantom with realistic lesion characteristics and respiratory motion.
- The 4D observer's performance was compared against conventional 3D observers (nongated and motion-corrected) using signal-to-noise ratio (SNR) measurements across various lesion and motion conditions.
Main Results:
- The 4D numerical observer demonstrated a significant average improvement in detection SNR of 48.6% compared to the nongated 3D method (p < 0.005).
- The 4D method showed a 13.8% higher relative SNR gain (Gain4D=1.49) than the motion-corrected 3D method (Gain3D=1.31) (p < 0.02).
- Superior performance was observed for lesions with high motion amplitude, and the 4D observer maintained linearity under motion, outperforming 3D methods across diverse conditions.
Conclusions:
- The proposed 4D numerical observer offers a substantial improvement in lesion detectability, providing a more reliable objective assessment of respiratory-gated PET benefits.
- This 4D approach can serve as an upper bound for evaluating the efficacy of motion correction techniques in PET imaging.
- Future validation on clinical data is planned to confirm the utility of the 4D observer in pulmonary oncology detection tasks.
