DoseSave
DoseSave · Topics · What Is NPS? (Noise Power Spectrum)
Image Quality

What Is NPS? (Noise Power Spectrum)

Standard deviation tells you 'how much' noise there is; but not 'how it looks'. Two images can have the same standard deviation yet one looks coarse and the other fine-grained. NPS (noise power spectrum) quantifies this 'noise texture' by distributing noise across spatial frequencies. What is NPS and why is it more informative than a single noise number? Concise.

Advanced · radiologists, technologists & physicists· ~7 min read· Sourced

We usually measure noise with a single number — the standard deviation. But that number is incomplete: it gives the amount of noise, not its texture. Just as MTF decomposes the signal by frequency, NPS (noise power spectrum) decomposes the noise by frequency. This article briefly explains what NPS is. For the signal side see MTF; for the efficiency that combines both, DQE.

What is NPS?

NPS (Noise Power Spectrum) — also known as the Wiener spectrum — is a function showing how noise is distributed across spatial frequencies.1 What MTF does for signal, NPS does for noise: a rich, frequency-dependent description. The key link: the integral of NPS over all frequencies equals the noise variance (σ²).1 So the standard deviation is just the total area under the NPS; NPS gives that total together with how it spreads across frequencies.

Noise texture

This is where NPS shows its power. Bushberg shows two images with the same standard deviation but different appearance: in one the noise is coarse (low-frequency-weighted), in the other fine-grained (high-frequency-weighted).1 A single number (standard deviation) cannot tell them apart; NPS distinguishes them by the shape of the curve. This matters especially in iterative reconstruction: these algorithms reduce the amount of noise while also changing its texture, sometimes creating a "plastic/artificial" look — a change only NPS captures.

Same σ · different texture → different NPScoarse texture (same σ)fine texture (same σ)NPSfrequency →coarsefine
The two images have the same standard deviation (equal total noise) but different texture: coarse noise concentrates at low frequency, fine noise at high frequency. The NPS curve shows this difference; a single σ cannot.1

Why it matters

Because the texture of noise affects visibility. Noise concentrated near a lesion's size (frequency) can mask it, while noise at a very different frequency is less disturbing. So modern detector and reconstruction evaluations do not rely on standard deviation alone; they use NPS (together with MTF). The two combine in DQE to define a detector's dose efficiency.

In a nutshell
NPS = the distribution of noise across frequencies (the Wiener spectrum). Its integral = the variance (σ²). The same σ can give different textures; NPS distinguishes them. Signal side → MTF; efficiency → DQE.
Related articles
What Is MTF? · What Is DQE? · Dose and Noise (√N) · Image Quality

References

  1. Bushberg JT, Seibert JA, Leidholdt EM, Boone JM. The Essential Physics of Medical Imaging, 3rd ed. Lippincott Williams & Wilkins, 2011. §4.6 (Noise Texture: The Noise Power Spectrum): NPS, gürültünün frekansa bağlı dağılımıdır; tüm frekanslar üzerindeki integrali varyansa (σ²) eşittir; aynı standart sapmaya sahip iki görüntü farklı gürültü dokusuna sahip olabilir (Şekil 4-27, s.86). Sayfa numaraları bu baskıya aittir.
  2. İlişkili: MTF Nedir? · DQE Nedir? · Doz ve Gürültü (√N) · Görüntü Kalitesi
Note: This content is for education; for clinical decisions or regulatory compliance, consult a qualified medical physicist and current regulations.

← All articles