Digital Subtraction Angiography (DSA)

subtraction of post-opacification image (live) from pre-opacification image (mask)

I2(live) = I1(mask).; μc, qc and tc = mass attenuation coefficient, concentration and thickness of contrast

D(subtracted image) = I1-I2 =, hence will contain artefacts from details under and overlying vessels from mask

logarithmic processing – reduces overlying artefacts by natural log I before subtraction so Dlog = μ

densitometric analysis based on calculates percent stenosis, LVEF, but assuming monoenergetic XR and no scatter

compression of grey scale from Dlog so multiplication with scaling factor applied to the transformed pixel

Dlog reduces with higher SPR (scatter-to-primary ratio); SPR usually > 1 without grid so air gap or grid will improve contrast of opacified vessels

improved scatter correction techniques use narrow fan beams, computerised scatter corrections


images are different (due to contrast), hence when they are added/subtracted, SNR reduces by √2

total noise (increased ~40% from subtraction process) = quantum noise + system noise (mostly from video camera)

when quantum noise is dominant source, SNR ; DII = absorbed D in air at II input phosphor

also improved by equalisation of transmittance throughout image (bolus materials over high transmittance regions), XR energy slightly above iodine K-edge (33keV)

when system noise is dominant source, increasing exposure is unlikely to decrease noise

imaging averaging (using image ALU) from sequence of images reduces noise by ; N = number of averaged images

temporal/recursive filtration involves exponentially-weighted moving average of images with SNR improvement

exposure can be low continuous (1-3mA) with digital noise reduction techniques applied; or intense pulsed exposures (100-1000mA) to produce high SNR images


video signal fed to digital image processor for manipulation, storage, display and control of video camera and high V generator (for pulsed exposure mode)

variable optical aperture between XII and video camera to prevent saturation of video camera (wide for fluoroscopic exposures, narrow for DSA exposures)

generators must produces exposures up to 100kVp, 1000mA with very short exposure times; also capable of low continuous exposures

XRT must have high heat capacity

Video Camera

electron guns based on diode configurations allowing large currents (thus lower noise)

SNR of 60dB (1000:1) required

good temporal resolution, hence low persistence (eg PbO)

  • specialised scanning mode with pulsed progressive readout (PPR) scan white target is blanked followed by a scrub frame
  • independence from exposure period allows exposure pulses of variable length
  • max frame rate ½ of continuous interlaced raster scanning

high resolution from 1049 or 2099 line Plumbicon tubes and CCD cameras producing 10242 or 20482 10 bit images at 25 or 7.5fps

Image Processing

video signals digitised (ADC) then passed through ILUT (input look-up table) for logarithmic transformation before or after storage

three image memories for mask, live and subtracted image

ALU (arithmetic/logic unit) performs image subtraction and averaging, with resulting image fed along feedback path to storage

array processor used for spatial enhancement and re-registration

OLUT (output look-up table) enhances contrast before DAC (digital to analogue conversion) for viewing

contrast enhancement varies window and level via OLUT, with null results from subtractions stored as mid-grey so positive and negative results can be visualised

time-interval-difference imaging periodically updates mask for fast-moving events

re-masking involves interactively choosing suitable mask to reduce motion artefacts

re-registration = pixel shifting spatially shifts mask relative to live to reduce motion artefacts

simultaneous imaging of arteries and veins by subtracting from each other to produce white/dark vessels

road-mapping – image at peak opacification used as mask for advancement of wire/catheter

land-marking – fraction (90%) of live subtracted from mask to leave some anatomical landmarks

bolus-chasing – contrast automatically tracked by moving table and/or XRT/XII

rotational angiography – C-arm rotates 10-30°/s to generate perceived 3D effect

volume tomographic angiography – C-arm used similar to CT for volume reconstruction

spatial enhancement = edge enhancement and smoothing

quantification – geometric analysis (number of pixels measured with calibration to convert to units) and densitometric analysis (computes pixel or mean pixel values)

  • measuring artefacts caused by spatial distortion from XII, scatter, veiling glare (scatter/glare correction used), relies on accurate determination of edges

temporal filtration = noise reduction = frame integration

conventional DSA uses many images (eg waiting for peak opacification, increasing patient dose) that does not contribute to production of diagnostic image; temporal filtration processes the images obtained over time to reduce noise

integrated mask-mode DSA – adding images prior to contrast to form mask, and peak-opacification images to form the live so less dose is wasted but instead contribute to lower noise

matched filtration utilises all images taken to produce dilution curves (for a certain region of the blood vessel) via densitometric analysis software to define weighting factors that are multiplied to each image before adding together; relatively high SNR

recursive filtration = exponentially-weighted moving average = fluoro noise reduction – integrates images as they are acquired with more recent images having a higher weighting factor

Dose Reduction Techniques

rotate tube and XII through 180° during prolonged procedures

pulsed fluoroscopy instead of continuous, reducing D by 90%

added filtration (0.1-0.3mm Cu added)

ROI filter = XR fovea (less filtration at central region)

circular collimators based on adjustable multi-leaf absorbers

radiation-free collimator positioning (storing image in memory or light beam diaphragms)

digital image storage – last image hold (LIH), reference image, image browsing