Three lesions (contrast 10%, 25%, 50%) in a disc phantom are imaged with Poisson counts. In the first step total counts rise from 10³ to 10⁶, the image sharpens and the measured relative noise follows 1/√n. In the second step each lesion's CNR at the chosen counts is compared with the Rose criterion.
Step 1/2
Counts and noise
Photons arrive as a Poisson process: in a pixel with n counts the standard deviation is √n and the relative noise 1/√n. As total counts rise the image stops looking grainy; absolute noise increases but relative noise falls and SNR = √n rises. Four times the counts halves the noise. The value measured in the background follows the theoretical curve.
Key point: Poisson noise dominates in nuclear medicine; relative noise falls as 1/√N and SNR rises as √N; detectability depends on CNR, and a low-contrast lesion needs many more counts.
Source: book (σ = √N, relative noise 1/√N, SNR = √N, CNR definition, Rose criterion CNR ≥ 3-5). Model: 48 × 48 pixel disc, lesions 7 pixels across; counts from seeded Poisson samples; CNR per pixel (lesion area not included, so the true detection threshold may be at lower counts).
How to use: a single click on the drawing pauses and resumes; in the second step ↑ ↓ or the buttons below change the total counts; ← → change steps.
Counting statistics and noise · Open in Nuclear Medicine Explained ↗