Noise, information that is not part of a desired signal, is present in all electronic systems, and originates from a number of sources including electronic interference. It appears as an irregular granular pattern in all images and degrades image information. It may be inapparent or render images non-diagnostic, depending on the severity. Noise should not be confused with other artifacts, which are less random and should be repeatable in theory, although noise is itself an artifact.
Noise in plain film depends on the number of discrete x-ray photons reaching the detector. The quantum noise (quantum mottle), structural noise and electronic noises are the main sources. Noise in plain radiography can be decreased by increasing the mAs which increases the number of photons. If we increase the dose by N times to the patient, noise decreases by square root N times.
Noise (quantum noise) in CT depends on the number of discrete x-ray photons reaching the detector. See the separate article on noise (CT).
The source of the electronic noise in CT is a combination of the noise of the detector system and the reconstruction kernel (sharper kernels give noisier images). Noise in a CT scan image can be decreased by increasing the mAs, by increasing the tube current or changing filters during reconstruction.
Source of noise
The main source of noise in the image is the patient's body (radiofrequency emissions due to thermal motion).The whole measurement chain of the MRI scanner (coils, electronics) also contributes to the noise.
This noise corrupts the signal coming from the transverse magnetization variations of the intentionally excited spins (on the selected slice plane).
Reducing MRI noise
While we can never eliminate noise from imaging, the following factors can be modified to reduce it when performing MRI:
- use the correct coil and ensure that it is well-tuned
- use a coarse matrix
- use a large field of view (FOV)
- select thick slices
- use narrow bandwidth
- use as many NEX as possible
Noise in an MRI image can be quantified from the standard deviation of a homogeneous area such as the background area or from multiple acquisitions (NEX>2) 1-3.
- 1. Sijbers J, Poot D, Den Dekker AJ et-al. Automatic estimation of the noise variance from the histogram of a magnetic resonance image. Phys Med Biol. 2007;52 (5): 1335-48. Phys Med Biol (abstract) - doi:10.1088/0031-9155/52/5/009 - Pubmed citation
- 2. Sijbers J, den Dekker AJ, Van Audekerke J et-al. Estimation of the noise in magnitude MR images. Magn Reson Imaging. 1998;16 (1): 87-90. Pubmed citation
- 3. Rajan J, Poot D, Juntu J et-al. Noise measurement from magnitude MRI using local estimates of variance and skewness. Phys Med Biol. 2010;55 (16): N441-9. Phys Med Biol (abstract) - doi:10.1088/0031-9155/55/16/N02 - Pubmed citation