CardioRender · Dr PDR's Teaching Series

C-arm & fluoro

positioning trainer for the table side one patient's CT · segmented chambers · registered coronaries
ANT
POST
L
R
N
Table-side module
CraCauRAOLAO
C-arm
HeadFeetRL
Table
TableDet

Rate control, as at the table: the further you push, the faster it moves; let go and it stops. The table stick moves the table — push toward the head and the patient slides headward, so the field falls lower on the chest. Hold an arrow to raise or lower the table or the detector.

Exposure — the pedals

Contrast goes in from the catheter tip and travels down the artery at the speed of coronary flow — mostly in diastole — reaching the distal branches in about a second, then clears from the ostium on. Inject on the deck: a tap gives a 2-second injection, holding injects for as long as you hold. A cine run injects by itself a moment after it starts.

Fluoro
Level
Cine
Scatter — what reaches you

Field of view — zoom

Zooming in magnifies the picture; the system raises the dose per frame to hold the noise, so the skin dose rate rises while the irradiated area — and the scatter — shrinks. Watch the Dose tab.

Collimation — set on the last image, no exposure

Drag an amber handle to bring a pair of shutters in. Drag the lower violet handle on either side to slide a wedge in; the upper one turns it. The same lines show on the live image while you work.

Projection

Drag on the image to swing the C-arm: sideways for LAO and RAO, up and down for cranial and caudal. The C-arm joystick above does the same at a controlled speed.

Display
Heartbeat

The coronaries ride on a beating heart: the base and the AV grooves travel furthest, so the RCA and circumflex move most and the apical LAD least. At 7.5 pulses a second a heart at 70 beats a minute is shown in about six steps per beat; at 30 in about twenty-five. Faster rates are why cine runs at 15 or 30.

Respiration — the heart rides on the diaphragm

Catheter — which ostium is engaged

Occlusion — contrast stops at the lesion
Stenosis — the view decides how tight it reads

Image processing — what the console does to the signal

Raw is the projection as the detector measures it: a dark heart in bright lung. Clinical is what a console shows: large differences compressed toward one light grey, vessels and edges enhanced. Every real angiogram you have seen was processed like this.

Beam & detector
Catheter — drive it yourself

Hold a button, or use the arrow keys: up and down push and pull, left and right turn. The catheter is an elastic rod with the maker’s curves, in this patient’s aorta: where its tip goes is worked out, not drawn.

Positioning drills — for the table side
Reading the view
Ventriculography — LV and RV

A pigtail sits in the chamber and 30 mL goes in over a few beats. Tap Inject on the deck. Each wall is tagged on the picture: a bright tag means that wall forms the border of the silhouette in this view, so its motion can be read; a faded tag means it faces the detector, overlapped by the cavity, and its motion is invisible here. Occlude a coronary on the Patient tab and its territory stops moving.

Radiation — experiments
How the image is formed — experiments

What this is. A coronary angiography simulator in which every pixel is computed: the beam is cast through a real adult chest CT and through a coronary tree grown on that patient’s own heart, built as closed tubes, and the detector sees I = I₀·exp(−ΣμL) along each ray. Nothing is drawn; the picture follows from the geometry. It has two jobs. For the table side — the Drive tab — it is a table-side module with rate-controlled sticks, table and detector height, and drills in isocentring, centring and collimating a patient. For the reader — Views and Ventricle — it shows why each standard projection exists, what foreshortening does to a lesion, and which wall of the LV or RV a ventriculogram can and cannot read in a given view. Radiation and Image are for everyone in the room: experiments on how the table, the detector, zoom, collimation, angulation and frame rate change the dose and the scatter, and on how the image is formed.

The data. The chest, the heart and the great vessels are one patient’s: a contrast-enhanced chest CT of a normal heart (256-slice scanner, 0.9 mm slices, 0.68 mm pixels, not ECG-gated), segmented with TotalSegmentator (lungs, ribs, spine, sternum, costal cartilages, liver, aorta, trachea, oesophagus, the great veins and arch branches) and with the MONAI whole-body cardiac model (LV and RV blood pools, LV myocardium, the atria and the pulmonary artery), with the RV free wall and atrial walls grown from the CT density on those chambers. Every label is a closed, watertight surface, and the heart region is carried at 1 mm with its Hounsfield values. The chest attenuation is built from the labels at 1.5 mm (air, lung, soft tissue, cortical shell and cancellous bone) and, within the heart region, from the scan’s own densities with the opacified blood taken back to plain blood. The ventriculograms use the opposite: each chamber’s own contrast, voxel by voxel, so the papillary muscles and trabeculae are the patient’s. The coronary arteries are the one thing the package lacks (the coronary model is licensed, and a scan that is not ECG-gated blurs them past their first centimetres), so the tree is grown on this heart. The main stems run in its own grooves, found from the labels: the left main from its ostium in the left sinus to the bifurcation, the LAD down the anterior interventricular groove and round the apex, the circumflex and the RCA round the mitral and tricuspid rings (each groove fitted as a smooth ring), the PDA down the posterior groove from the crux and the posterolateral branch on past it. The branches are walked over the epicardium toward their territories: diagonals to the lateral apex, marginals down the lateral wall, posterolateral branches over the inferolateral wall, the conus, the sinus-node artery and the RV branches by the shortest paths over the surface; the septal perforators run inside the septum, held on its mid-surface. Calibres follow Murray’s law at every branch, with a slow taper. The heart is right-dominant. Nothing that identified the patient is carried.

Gantry and naming. A single-plane C-arm keeps the tube under the table and the panel over the chest, so the beam always runs posterior to anterior; angiography names the projection for the detector, which is why the frontal view is called AP. LAO swings the panel to the patient’s left, cranial tilts it toward the head. The default perspective puts the eye at the tube and mirrors the image for display, as a radiograph is read from the detector side, so the magnification is the lab’s: SID 100 cm over SOD 70 gives 1.43× at the isocentre, more for whatever lies nearer the tube. Raising the detector lengthens the SID and magnifies; lowering the table brings the heart toward the tube and does the same. A detector-side view on the Image tab reverses the depth-dependent magnification for readers who find it easier; every left–right relation and every swing is unchanged by it, and the burned-in numbers always give the true X-ray values. When the table is centred on the left coronary the median plane sits a few centimetres off the central ray, so the spine and sternum superimpose not in AP but a few degrees to one side; the beam cues on the Image tab name the angle.

The image chain. Path lengths are accumulated on the GPU as signed chords through closed surfaces, iodine, bone and soft tissue each in its own channel. Bone carries its measured density and a beam-hardening term (L/(1 + L/6 cm)). An automatic brightness loop reduces a 64×64 render to one texel, weighted toward the centre, and drives the tube until the set point is met, settling over 0.4 s; the tube output it asks for is burned into every frame and scales the dose. Scatter is added as a pedestal proportional to the incident field (scatter-to-primary 0.35 by default, about what a grid leaves, shrinking with the square root of the collimated area) and veiling glare as a few percent of the low-passed primary. Quantum mottle is one Gaussian grain per rendered pixel with the photon count set by the dose per pulse, summed over the last six pulses with the console’s recursive temporal filter (time constant 0.12 s), so higher pulse rates look cleaner and cost more. Focal-spot blur is applied to bone at its own depth, read from the first-moment buffer. The console then takes the log, splits the picture into size bands and compresses the large ones (harmonisation) while enhancing the small, cored at the pixel’s own noise, and displays density through a window with a soft shoulder. Every slider on the Image tab moves one of these links and nothing else.

Stenosis. A lesion is a Gaussian hourglass in the tube’s own rings, so the residual lumen is a real closed solid the projection integrates through. The readout does not assume what you will see: it projects the residual lumen through the actual source and detector axes, bins the silhouette along the projected axis and reports the narrowest width against the widest — the diameter stenosis a reader would measure on that frame — then searches the gantry for the angle at which it reads truest. As a lesion turns toward the beam its ends stack over its waist and the number falls; that gap is why an operator hunts for a working view before calling a number.

Contrast. An injection sends a rounded front down the tree at coronary velocity — about 125 mm/s on the left and 200 on the right, faster in diastole and nearly stopped in systole — so the RCA fills in about 0.7 s, the LCx in 0.85 and the LAD in just over a second, in the order the TIMI frame counts run; then it clears from the ostium on. A guide catheter of the chosen shape sits in the ostium, its shape the equilibrium of an elastic rod in this aorta, and reads half as dark without contrast in it.

Ventriculography. The LV (100 mL) and RV (122 mL) are the scan’s own blood pools: the ray-marcher weights each labelled voxel’s opacification by the injection, so the filling defects of the papillary muscles and trabeculae, the mitral plane and the outflow tracts are the patient’s. A pigtail enters from the femoral artery through the aortic valve, or up the IVC through the tricuspid valve, along centrelines skeletonised from the labels, and 30 mL fills the chamber over a few beats and washes out; what the LV ejects runs out into the real aortic root and ascending aorta, what the RV ejects into the pulmonary trunk. The cavities beat with the cavity gains and the walls with a third of them — the endocardium sweeps in and the wall thickens, the base descends about a centimetre, the LV ejects about 60% and the RV about 50% — and a wall can be made akinetic by hand or by occluding its artery on the Patient tab, which drops the ejection fraction accordingly. Each wall of the blood-pool surface is classed from the groove geometry, and its tag is bright when enough of it forms the border of the silhouette, where its motion can be read, and faded when it faces the beam and is overlapped by the cavity: RAO 30 reads the anterior wall, apex and inferior wall; LAO 60 with cranial reads the septum and the lateral wall; the RV’s inflow and outflow open in RAO and its free wall against the septum in LAO. The rule is geometry, not a table, so it holds at every angle.

The dose model. Air kerma is anchored at 0.011 mGy per normal fluoro pulse (half for Low, double for High) and 0.10 mGy per cine frame at the interventional reference point, 15 cm from the isocentre toward the tube, for an average patient in AP; it is scaled by the tube output the brightness loop asks for, by the square of the SID, by the inverse of the field size and by the detector dose. The dose–area product multiplies that by the field at the reference point after collimation. The skin dose follows the tube–skin distance along the central ray with the inverse square, so table height changes it and nothing else. Scatter to the staff is 2 µGy per Gy·cm² of DAP at one metre side-on, up to three times that back toward the tube from where the beam enters, half that beyond the patient, falling with the square of the distance; ceiling shield ×0.1 at the eyes, table skirt ×0.1 at the legs, apron and collar ×0.05, lead glasses ×0.35. The counters start at your first touch. These are illustrative, order-of-magnitude figures for teaching; only a dosimeter measures what you receive.

Honest limits. No collaterals, no myocardial blush, no reflux into the root. Flow is the same in every branch. A branch’s first millimetre lies inside its parent and is counted twice, so a branch origin reads a shade dark. The coronary tree is a model grown on this heart, not segmented from it: its course is this heart’s grooves and surface, its branching the usual right-dominant pattern, and no one has exactly this tree. The chest outline is the convex hull of the segmented structures grown by 14 mm, not the skin itself, so the breasts and the arms are not there. The scan is not ECG-gated: the chambers are an average over the cycle and beat only as the model makes them, and breathing moves the soft tissue without the ribs’ lift. The scan covers 28 cm of the chest; beyond it the edge slices are carried on. Catheter shapes leave out friction and twist. Illustrative, not diagnostic.

Attribution. Chest and heart: an anonymised contrast chest CT of a normal heart, segmented with TotalSegmentator 2.18 (Wasserthal et al., Radiology: Artificial Intelligence 2023) and the MONAI model zoo whole-body CT segmentation (SegResNet). Coronary arteries: grown on this heart’s epicardium and grooves. Rendering: three.js (MIT). Catheter geometry from the makers’ curve definitions. © Dr Dinesh Raja Palanisamy · CardioRender · Dr PDR’s Teaching Series. For education and training only; not a diagnostic or clinical decision tool.

Education and training only · not a diagnostic or clinical decision tool · © Dr Dinesh Raja Palanisamy · CardioRender