Authors
Chenyang Han, Andrew T Trout, Andi Li, Joseph G Meier, Cara E Morin, Adam F Prasanphanich, Susan E Sharp, Christopher G Anton, Jing Tang
Published in
NeuroImage. Pages 122328. Nov 03, 2026. Epub Nov 03, 2026.
Abstract
To generate dedicated brain PET-quality images from routine whole-body PET acquisitions by robustly enhancing the brain region using a noise-aware deep image prior (NA-DIP) approach.
NA-DIP integrates anatomy-guided priors derived from patient-specific μ-maps or MR images, a regularization strategy for enhanced denoising, and a variance-based early stopping criterion that adaptively monitors noise without requiring prior noise assumptions. Using noisy PET images as self-supervised targets, the method was first evaluated using simulation data spanning three noise levels to investigate the trade-off between noise suppression and bias. It was then applied to brain images extracted from 90-s-per-bed-position whole-body acquisitions from 15 pediatric FDG-PET studies spanning a broad range of patient demographics, physiological variability, and image noise levels. Using the corresponding 360-s-per-bed-position dedicated brain scans as the reference, NA-DIP-enhanced images were evaluated for noise reduction, standardized uptake value (SUV) accuracy, and image quality in a blinded reader study.
NA-DIP demonstrated a superior noise-bias trade-off compared with Gaussian filtering and conventional DIP stopping strategies, particularly under high-noise conditions. The inclusion of MR or μ-map guidance further improved performance. Across pediatric patients, NA-DIP-enhanced 90-s images exhibited lower noise than the corresponding 360-s dedicated brain images while showing modest variability in SUV measurements. Blinded readers rated the NA-DIP-enhanced images as comparable to or better than dedicated brain images in most cases.
NA-DIP enhanced brain PET images from whole-body acquisitions demonstrated good quantitative agreement and image quality comparable to or exceeding the reference, supporting the feasibility of eliminating separate dedicated brain scans.
PMID:
42977202
Bibliographic data and abstract were imported from PubMed on 04 Nov 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 10
- Comments 0