Image quality and noise in DSA – Learn Radiology Physics

Unlike static radiography, DSA image quality is influenced by both detector performance and the mathematical effects of subtraction, which amplify noise and magnify the impact of motion and registration errors.

Understanding these relationships allows optimisation of dose, frame rate, and processing.

Determinants of image quality in DSA (summarised)

Factor	Influence	Optimisation method
Detector DQE	Higher DQE improves SNR for a given dose	Use flat-panel detectors with high conversion efficiency
Photon fluence (dose per frame)	Determines quantum noise level	Optimise pulse mA and exposure time
Frame rate	Affects temporal resolution and total dose	Match to vascular flow speed
Subtraction accuracy	Imperfect alignment produces residual artefacts	Pixel shift and remasking
Temporal filtering	Improves SNR at cost of temporal blur	Adjust weighting to flow rate
Pulse width	Shorter pulses reduce motion blur	Requires higher instantaneous output

Each parameter is a balance between image clarity, temporal accuracy, and radiation dose.

Signal-to-Noise Ratio (SNR) in DSA

Image noise arises primarily from quantum noise and detector electronic noise, but in DSA, additional noise is introduced through subtraction.

For uncorrelated images:

σ²_DSA = σ²_mask + σ²_live

If both frames have similar noise (σ\sigmaσ):

σ_DSA = √2 σ

Thus, the noise in a subtracted image is approximately 1.4× higher than in a single image, and SNR decreases accordingly.

SNR increases with the square root of dose:

SNR ∝ √Dose

Because DSA introduces √2 more noise, achieving the same SNR as a non-subtracted image requires approximately twice the dose per frame.

Optimisation therefore targets adequate, not maximal, SNR, relying on filtering and averaging to enhance appearance without unnecessary exposure.

Temporal and spatial resolution

Temporal resolution

Determined by frame rate and pulse width.
High frame rates (15–30 fps) improve depiction of rapid arterial filling but increase dose and noise.
Lower frame rates (3–7.5 fps) are used for slower venous or peripheral studies.
Temporal filtering smooths noise but can blur fast-moving boluses.

Spatial resolution

Limited by detector pixel size (150–200 µm typical).
High-contrast edges (vessels) benefit from the inherently high MTF of modern flat-panel detectors.
Geometric unsharpness is minimised by using a small focal spot and keeping the object-to-detector distance short.

Contrast resolution and CNR

Contrast-to-noise ratio (CNR) is more relevant than SNR in DSA, as diagnostic visibility depends on iodine signal strength relative to background noise:

CNR = (∣μ_iodine−μ_tissue∣) / σ

Optimisation

Operate at 60–80 kVp so that a large part of the spectrum lies just above the iodine K-edge (33 keV).
Maintain consistent exposure across frames for accurate subtraction.
Use logarithmic subtraction, which linearises contrast with iodine concentration.
Apply temporal averaging or recursive filtering to stabilise vessel signal.

Noise sources and propagation

Noise Source	Description	Impact / Correction
Quantum noise	Random photon arrival variation	Dominant; improved by higher dose or filtering
Electronic noise	Detector readout or amplifier noise	Negligible in modern systems
Subtraction noise	Combined mask + live noise	Increases total variance by √2
Motion artefact noise	Residual edges from misregistration	Minimise via immobilisation, pixel shift
Lag / ghosting	Detector afterglow	Correct by calibration and short integration times

Noise propagation is cumulative; effective dose management and filtering maintain SNR without compromising temporal accuracy.

Optimising image quality

Exposure parameters

Use short pulses (5–10 ms) for motion freeze.
Set mA high enough to ensure adequate photon fluence for each frame.
kVp of 70–80 provides optimal iodine contrast and penetration.

Processing techniques

Apply recursive filtering for stable background noise reduction.
Use pixel shift and remasking to correct motion artefacts.
Limit edge enhancement to avoid amplifying noise.

System design

High DQE detectors and efficient anti-scatter grids maximise dose utilisation.
Automated gain calibration maintains uniform subtraction.

Key takeaways and exam tips:

Subtraction doubles image noise (√2 increase in variance).
High DQE and adequate per-frame dose are essential to maintain SNR.
Temporal resolution depends on frame rate and pulse width; balance against dose.
Contrast resolution peaks near 33 keV (iodine K-edge).
Recursive filtering (frame averaging) and pixel shift improve noise and registration without re-exposure.
Common exam question: “Explain how noise is propagated in DSA and how image quality can be optimised without increasing dose.”

Um, what!? Can you believe it? You’ve reached the end of the X-ray physics notes!

Congratulations 🎉

That’s no small achievement.

You’ve worked through some of the most conceptually demanding areas of radiology physics, from the fundamentals of electromagnetic radiation all the way to the complexities of digital subtraction angiography.

That takes commitment, consistency, and real intellectual effort.

Take a moment to recognise what you’ve accomplished.

These notes represent not just hours of study, but the development of a deeper understanding of how medical imaging truly works. Knowledge that will serve you throughout your career.

Thank you for trusting me to guide you through it, and for investing the time and focus needed to get here.

You should be proud of what you’ve achieved.

Now, as you prepare for your exams and your future practice, remember, mastering physics isn’t just about passing an exam, it’s about becoming a more confident, safer, and more capable clinician.

Congratulations again, and well done!