Image processing in DSA

Digital subtraction angiography (DSA) depends heavily on post-processing.

Even with optimal acquisition, residual artefacts, noise, and motion can obscure vascular detail.

Modern DSA systems therefore apply a suite of processing techniques (both automated and operator-controlled) to enhance diagnostic clarity while maintaining quantitative accuracy.

In depth knowledge of these techniques isn’t necessarily required for most radiology physics exams. These, at most, represent one or two marks in an exam (from past exams, you never know what will crop up in future exams though 😅). But, I think it’s worthwhile having at least some basic knowledge of these concepts.

Objectives of post-processing

• Optimise vessel visibility by enhancing contrast-to-noise ratio (CNR).
• Correct misregistration artefacts from motion between mask and live images.
• Suppress quantum noise without degrading temporal or spatial detail.
• Improve workflow through real-time correction and display during procedures.

These techniques collectively reduce the need for repeat runs, lowering both dose and contrast usage.

Temporal filtering and noise reduction

Quantum noise limits SNR, particularly at the low per-frame exposures used in serial DSA runs. Temporal filtering combines data from consecutive frames to smooth random fluctuations.

Temporal filtering increases effective SNR but slightly reduces temporal resolution; settings are adjusted to match the vascular region (e.g. tighter filtering for steady abdominal flow, looser for cerebral circulation).

Pixel shift

Small patient or table movements between the mask and live images lead to misregistration artefacts, appearing as double edges or ghosting.

Pixel shift corrects this by digitally translating the mask image by integer or fractional pixel increments to best align with the live frame. The system measures correlation across frames and applies sub-pixel adjustments in both x and y axes.

This restores accurate subtraction without reacquiring data.

Remasking

If the initial mask becomes contaminated, for example, by early contrast arrival or motion, a new mask can be selected from within the live series.

This process is known as remasking.

• The operator designates a later frame (pre-arterial or early phase) as the replacement mask.
• The sequence is re-subtracted using this new reference, eliminating artefacts from the compromised original.
• Particularly useful in body angiography, where respiration or peristalsis can shift anatomy during acquisition.

Road-mapping

Road-mapping provides dynamic guidance during interventional procedures.

A subtracted DSA image (showing vessel outlines) is superimposed on continuous fluoroscopy, forming a static “road map.”

As catheters or guidewires move under live fluoroscopy, their position is seen relative to the fixed vessel pathway.

Feature	Function
Static subtracted background	Provides vascular reference
Continuous low-dose fluoroscopy	Displays device movement
Automatic registration	Keeps overlay aligned during table movement
Benefit	Reduces need for repeated contrast injections and full DSA runs

Road-mapping is especially valuable in neuro- and peripheral interventions, where navigation precision is critical.

Energy subtraction (dual-energy DSA)

Not all DSA techniques require a mask image. Dual energy DSA relies on the fact that tissues attenuate X-rays differently at different X-ray energies. We cover this in great detail in the CT physics mastery course.

Briefly, while standard DSA removes background structures temporally, dual-energy DSA achieves subtraction using energy discrimination.

Principle

Two images are acquired at photon energies just below and above the iodine K-edge (33 keV). Because iodine attenuation changes sharply across the K-edge, the difference between these images isolates iodine signal independent of motion.

I_sub = I_high − k × I_low

where k is a weighting constant adjusted to equalise soft-tissue contrast.

Advantages

• Less sensitive to motion than temporal subtraction.
• Allows imaging even when patient movement or pulsation prevents a stable mask.
• Provides quantitative iodine mapping.

Limitations

• Requires rapid kVp switching or dual-beam generation.
• Higher dose and greater system complexity.
• Energy separation limited by available X-ray spectra.

Key takeaways and exam tips:

• DSA image quality depends as much on post-processing as on acquisition.
• Temporal filtering improves SNR at the expense of temporal resolution.
• Pixel shift and remasking correct alignment errors and motion artefacts.
• Road-mapping overlays a static vascular map for interventional navigation.
• Dual-energy subtraction isolates iodine using spectral rather than temporal differences.
• Over-processing can introduce artefacts or misrepresent true vessel density.
• Common exam question: “Describe the post-processing techniques used in digital subtraction angiography and their roles in improving image quality.”

Up next

Next, we’ll move on to Artefacts and Limitations in DSA, describing common sources of image degradation such as motion, misregistration, contrast contamination, and detector artefacts.