Impact of Acquisition Geometry, Image Processing, and Patient Size on Lesion Detection in Whole-Body 18F-FDG PET

Georges El Fakhri, Pedro A. Santos, Ramsey D. Badawi, C.h. Holdsworth, A. D. Van Den Abbeele, Marie Foley KijewskiJournal of Nuclear Medicine 2007 · DOI

Abstract

In whole-body $^{18}$F-FDG PET imaging, the choice between 2D and 3D acquisition modes significantly impacts lesion detectability based on patient size. This study found that 3D imaging yields superior results for patients of normal size (BMI ≤ 31), while 2D imaging remains more effective for larger patients (BMI > 31) due to performance degradation in 3D mode. Furthermore, advanced reconstruction algorithms like Attenuation-Weighted OSEM (AWOSEM) and noise reduction techniques like Randoms Variance Reduction (RVR) provide substantial improvements in signal-to-noise ratios (SNR) across all patient sizes.

Introduction

Positron Emission Tomography (PET) scanners generally operate in two modes: 2D, which uses lead or tungsten "septa" (collars) to block out-of-plane radiation, and 3D, which removes these collars to maximize sensitivity. While 3D mode captures more signal, it also introduces significant noise from scattered radiation and random coincidences. This study aims to identify the "tipping point" where the noise inherent in 3D acquisition outweighs its sensitivity benefits, specifically examining the role of patient Body Mass Index (BMI).

Methodology: Hybrid Simulation and Numeric Observers

To evaluate performance with a known "gold standard," researchers employed a hybrid approach. They combined real clinical data from 32 patients with synthetic lesions generated from physical radioactive spheres (1.0, 1.3, and 1.6 cm). These synthetic lesions were mathematically inserted into the raw patient sinograms to create a controlled dataset for testing.

Methodology for generating hybrid datasets by inserting synthetic sphere sinograms into patient data.

Rather than relying solely on human interpretation, the study utilized a Channelized Hotelling Observer (CHO). This mathematical model mimics the human visual system's ability to detect signals against noisy backgrounds, providing a rigorous and objective metric for lesion detectability (CHO-SNR).

The BMI Tipping Point

The most critical finding of the research is the identification of BMI 31 as the crossover point for acquisition geometry. For patients with a BMI ≤ 31, the high sensitivity of 3D mode makes lesions significantly easier to detect ($P < 0.025$). However, as patient mass increases, the volume of scatter and random noise in 3D mode becomes overwhelming. For patients with a BMI > 31, the physical shielding provided by 2D septa becomes necessary to maintain image quality, making 2D the superior choice ($P < 0.05$).

Comparison of 2D and 3D AWOSEM reconstruction across different patient BMIs, showing the degradation of 3D performance in larger patients.

Algorithmic Improvements and Noise Reduction

The study also highlighted that software is just as important as hardware. Switching from standard OSEM to Attenuation-Weighted OSEM (AWOSEM) improved the SNR by approximately 40%. Additionally, implementing Randoms Variance Reduction (RVR) instead of standard Delayed Window Subtraction (DWS) provided a systematic boost in image quality across all patient sizes and acquisition modes.

Conclusion and Clinical Implications

This research provides a clear guideline for clinical practice: 3D PET is preferred for normal-sized patients, but 2D remains the gold standard for the obese. While newer technology like Time-of-Flight (TOF) may shift this BMI tipping point in the future, the fundamental relationship between patient mass, scatter noise, and detector geometry remains a cornerstone of nuclear medicine optimization.

Reference: El Fakhri G, Santos PA, Badawi RD, et al. Impact of Acquisition Geometry, Image Processing, and Patient Size on Lesion Detection in Whole-Body $^{18}$F-FDG PET. J Nucl Med. 2007;48:1951–1960.

Summary updated

Abstract

3D imaging yielded better lesion detectability than 2D (P < 0.025, 2-tailed paired t test) in patients of normal size (body mass index [BMI] < or = 31). However, 2D imaging yielded better lesion detectability than 3D in large patients (BMI > 31), as 3D performance deteriorated in large patients (P < 0.05). 2D and 3D yielded similar results for different lesion sizes. CHO-SNR were 40% greater for AWOSEM, FBP-ATT, and FBPNAT than for OSEM (P < 0.05), and AWOSEM yielded significantly better lesion detectability than did FBP. In all patients, RVR yielded a systematic improvement in CHO-SNR over DWS in both 2D and 3D. radicalNEC was characterized by a behavior similar to that of SNR(CHO) for the 3 different phantom sizes considered in this study.
BibTeX
@article{fakhri2007impact,
  author = {Georges El Fakhri and Pedro A. Santos and Ramsey D. Badawi and C.H. Holdsworth and A. D. Van Den Abbeele and Marie Foley Kijewski},
  title = {{Impact of Acquisition Geometry, Image Processing, and Patient Size on Lesion Detection in Whole-Body 18F-FDG PET}},
  journal = {Journal of Nuclear Medicine},
  year = {2007},
  doi = {10.2967/jnumed.108.007369},
}
Figure