Optimization of wavelet processing of dynamic PET data
Bibliographic record
Abstract
Several groups have reported that wavelet denoising can be used to increase the signal-to-noise (SNR) of dynamic PET studies (4D PET). But wavelet denoising can be applied in different ways and the process involves the arbitrary choice of strength parameters. Too much denoising results in a loss of resolution and biased concentration histories. The goal of this work is to study how to optimize wavelet denoising for neuroreceptor studies. Event by event Monte Carlo simulation was used to create an ensemble of 50 dynamic data sets. The simulation incorporates the geometry, electronic and detector characteristics of a commercial PET scanner (ECAT HR+) as well as the nuclear interactions necessary to describe the coincidence, random and scatter events. Temporal and spatial distribution of radioactivity was defined using a digital brain phantom and measured kinetic parameters. Sinograms for 70 minutes of data were computed to be similar to the noise level of 11C-raclopride in the human brain and then reconstructed using filtered back projection. 3D-wavelet denoising was applied to the ensemble of simulations. We studied denoising schemes that used either a fixed threshold for all wavelet coefficients or alternatives in which we varied thresholds for each sub-band. The wavelet transform of the mean of 50 fifty simulations was taken as the expected value. The RMS difference between the mean and noisy simulations was minimized by varying the thresholds. Noise was measured as the voxel-wise sample standard deviation for both conventional and denosied data sets. We also computed binding potential (BP) for the raw and processed data, with voxel-wise mean and standard deviation as end points. Bias was computed for both concentration and BP as the difference in means for the raw and denoised data set. Three-dimensional wavelet denosing with a single threshold yielded increased SNR, but only at the expense of significant bias. Optimization of sub-band thresholds yielded a 2x improvement in SNR for the 4D concentration data (See Fig. 1) and a 1.5x improvement for PB. More than 84% of voxels had bias less 3%. We conclude that optimized wavelet denoising can increase SNR in dynamic PET with low bias and minimal loss of resolution.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".