Rotational roadmapping: a new image-based navigation technique for the interventional room.
ABSTRACT For decades, conventional 2D-roadmaping has been the method of choice for image-based guidewire navigation during endovascular procedures. Only recently have 3D-roadmapping techniques become available that are based on the acquisition and reconstruction of a 3D image of the vascular tree. In this paper, we present a new image-based navigation technique called RoRo (Rotational Roadmapping) that eliminates the guess-work inherent to the conventional 2D method, but does not require a 3D image. Our preliminary clinical results show that there are situations in which RoRo is preferred over the existing two methods, thus demonstrating potential for filling a clinical niche and complementing the spectrum of available navigation tools.
- SourceAvailable from: Christian Unger[Show abstract] [Hide abstract]
ABSTRACT: 2D angiographic roadmapping is used frequently during image guided interventions to superimpose vessel structures onto currently acquired fluoroscopic images. While the fluoroscopic images, acquired with 12-15 frames per second, show patient bone anatomy as well as the current location of the inserted catheter, the roadmap delineates vessels to provide path information and to avoid accidental vessel wall punctures during catheter advancement.Computer Vision and Pattern Recognition, 2008. CVPR 2008. IEEE Conference on; 07/2008
N. Ayache, S. Ourselin, A. Maeder (Eds.): MICCAI 2007, Part II, LNCS 4792, pp. 636–643, 2007.
© Springer-Verlag Berlin Heidelberg 2007
Rotational Roadmapping: A New Image-Based
Navigation Technique for the Interventional Room
Markus Kukuk1,2 and Sandy Napel2
1 Siemens Medical Solutions, Forchheim, Germany
2 Stanford University, Department of Radiology, USA
Abstract. For decades, conventional 2D-roadmaping has been the method of
choice for image-based guidewire navigation during endovascular procedures.
Only recently have 3D-roadmapping techniques become available that are
based on the acquisition and reconstruction of a 3D image of the vascular tree.
In this paper, we present a new image-based navigation technique called RoRo
(Rotational Roadmapping) that eliminates the guess-work inherent to the con-
ventional 2D method, but does not require a 3D image. Our preliminary clinical
results show that there are situations in which RoRo is preferred over the exist-
ing two methods, thus demonstrating potential for filling a clinical niche and
complementing the spectrum of available navigation tools.
The number and breadth of minimally invasive, image-guided therapies is ever in-
creasing. Of particular interest are endovascular procedures, which allow minimally
invasive access to all areas of the human body through the vascular system as, for ex-
ample angioplasty, vascular stenting, embolization, chemoembolization, thrombolysis
and TIPS (Transjugular Intrahepatic Portosystemic Shunt). These procedures are typi-
cally performed in an interventional room using C-arm based X-ray imaging (see Fig.
1), together with the selective injection of contrast material for blood vessel opacifica-
tion. Common to all endovascular procedures is the navigation of a guidewire or
catheter through the vasculature to a target site. Depending on the degree of tortu-
ousity and structural complexity of the vascular tree, especially in diseased vascula-
ture, guidance (image- or sensor-based ) is often required for targeted steering.
Essentially unchanged from its first introduction in the early 1980s, a technique
called 2D-roadmapping  has long become clinical routine for image-based
guidewire navigation. The basic idea is to acquire an image of the vasculature of in-
terest and to store it as a “roadmap” image. Then, the guidewire or other instrument as
shown under live fluoroscopy is continuously superimposed onto the roadmap image,
thus visualizing the instrument with respect to the vasculature. For acquiring the
roadmap image, the interventionalist estimates how to position the C-arm for finding
a suitable working view, which is considered ideal if it shows the vessel bifurcation to
be negotiated perpendicular to the viewing direction, thus eliminating self occlusion.
However, the vessel tree is visible only after the injection of contrast media, which
Rotational Roadmapping 637
Fig. 1. Left: C-arm system “Siemens Axiom Artis dTA” (top) and its table side controls (bot-
tom). Arrows indicate the C-arm joystick (left) and RoRo joystick (right). Right: RoRo screen,
consisting of three columns: (a) image panel, (b) alignment signal, (c) control panel.
may or may not reveal the desired viewing angle. During the course of a lengthy in-
tervention, the individual injections of contrast material, which is toxic in large doses,
can add up to a significant amount due to the trial-and-error nature of the approach.
Only recently, a new image-based navigation technique has been introduced: 3D-
roadmapping . This technique is based on the acquisition and reconstruction of a
3D image (C-arm CT, CTA, MRA) of the vessel tree, which then serves as the road-
map image. CTA and MRA are acquired during an intra-venous contrast injection and
require registration with the C-arm system (2D/3D registration ). C-arm CT
images, acquired during a selective, intra-arterial injection during the procedure are
obtained from a rotational acquisition of projection images and the subsequent appli-
cation of a cone-beam reconstruction method . Following a one-time calibration
step, one can correctly render an image of the 3D volume as would be “seen” by any
given C-arm configuration. In other words, roadmap images are available from any
angle, thus eliminating the guess-work in finding an ideal working view. In situations
where a 3D angiographic acquisition is routinely performed for diagnostic purposes,
3D-roadmapping appears to be the perfect navigation tool at no extra “cost” regarding
dose or contrast agent.
However, there are clinical situations in which a 3D acquisition for the purpose of
navigation is either not justified or technically challenging. The following represents a
list of situations that may cause reconstruction artifacts and may therefore result in
compromised image quality: cardiac and respiratory motion during acquisition, inho-
mogenous flow of contrast material due to cardiac pulsation, high velocity flow of
contrast material, photon starvation along the shoulder axis and in the presence of in-
dwelling metal such as coils and stents. Furthermore, image quality is directly related
to the number of acquired projections and therefore a function of volume of contrast
In this paper we present a new image-based navigation technique that fits techni-
cally and clinically between the described 2D- and 3D roadmapping techniques:
638 M. Kukuk and S. Napel
Rotational Roadmapping (RoRo). RoRo is based on a single rotational acquisition of
multiple views (2.5D-roadmapping). Because 3D reconstructions are not performed,
acquisitions of any length and any number of projections are possible. Instead, the
projections are directly used for roadmap navigation.
RoRo can be regarded as a natural extension of the conventional 2D roadmapping
technique, as it allows rotating the C-arm during contrast injection, instead of keeping
it stationary. Thus, more information is acquired per unit of contrast agent, producing
a “rotatable roadmap” that shows the vessel tree from multiple views. After the acqui-
sition, each projection of the rotatable roadmap can be selected to serve as a roadmap
image for classic 2D-roadmapping navigation. The same rotatable roadmap can be
used again and again to find the ideal working view for each segment of the vessel
tree, thus rendering the injection of additional contrast media unnecessary. Further-
more, the vessel trajectories can be viewed in 3D, using stereographic visualization
(3D glasses) using suitably selected pairs of views, separated by a small angle .
We have developed a research prototype that implements the RoRo approach, with
software and hardware fully integrated into a C-arm system (Axiom Artis dTA, dBA,
Siemens Medical Solutions) that can be controlled during an intervention from the pa-
tient table side (see Fig. 1). We present first clinical results that demonstrate RoRo’s
potential for filling the gap between the existing methods.
2 Material and Methods
Although RoRo’s principal idea is straightforward, several challenges need to be ad-
dressed in order to provide a solution that meets the high demands of clinical, intraop-
erative software: seamless workflow integration, accurate C-arm/image alignment and
vessel tree/instrument visualization.
2.1 Clinical Workflow
1. Image acquisition (not part of RoRo)
2. Image transfer to RoRo
3. RoRo Phase 1: Find (suitable) working view
4. RoRo Phase 2: Align C-arm with selected working view
5. RoRo Phase 3: Roadmapping (Instrument guidance using selected working view)
6. If a different working view is needed, return to RoRo Phase 1
Image acquisition consists of a conventional rotational acquisition of any length and
angular increment. For example, in case of a rotational DSA (Digital Subtraction An-
giography) acquisition, two identical rotational runs are performed: the first without
(mask run) and the second with the injection of contrast media (fill run). After acqui-
sition, the study consisting of two runs of 2D projections is sent to the RoRo applica-
tion by means of a DICOM transfer.
Upon completion of the image transfer, RoRo computes a third “DSA run,” by
subtracting corresponding fill from mask images. RoRo then automatically enters
phase 1 by displaying the DSA run as a “rotatable roadmap.” For finding a suitable
working view, the C-arm and the rotatable roadmap are linked to each other in two
Rotational Roadmapping 639
Fig. 2. Interactive C-arm/image alignment. Left: C-arm joystick; Center: Alignment signal.
Right: C-arm’s Left/Right (top) and Head/Feet (bottom) plane, indicating the current C-arm po-
sition by an arrow and the currently displayed image by a box. Images were acquired every 1.5˚
(dots) along the Left/Right plane at 0˚ Head/Feet angulation. The deviation between the C-arm
position and the currently displayed image is indicated by the alignment signal. The top half
shows the alignment for the Left/Right plane to be within 1.0˚ (light blue) while the bottom half
shows the alignment for the Head/Feet plane to be within 0.5˚ (dark blue). The two arrows in
the alignment signal indicate the direction in which to move the joystick to improve alignment
for the respective plane: “upward” and “left” will move the C-arm closer to 93˚L/R and 0˚ H/F.
ways, as described in the next section. At this point, zoom formats and SID (Source
Image Distance) can be adjusted as needed.
Once the C-arm and the selected working view are aligned, RoRo enters Phase 3
“roadmapping.” By pressing the fluoro footswitch, live fluoroscopic images of the in-
struments being guided are acquired and overlaid onto the selected view of the vessel
tree, similar to the classic 2D roadmapping technique. At any time during Phase 3, the
user can return to Phase 1 for finding a new working view. This is done either by
moving the C-arm into a new position (roadmap follows C-arm), or by using the
RoRo-joystick to rotate the roadmap (C-arm follows roadmap). The RoRo joystick is
sterilized together with the standard controls by draping them in clear plastic sheets.
2.2 C-Arm/Image Alignment
One principal challenge with the RoRo approach is to provide for a fast and accurate
alignment of the C-arm with the currently displayed working view. After image
acquisition, vessel images are only available at discrete points along the acquisition
trajectory (see Fig. 2, right), while the C-arm operates in continuous space and can
therefore be moved in any position. Accurate roadmap visualization can only be pro-
vided if the C-arm is positioned close enough to the exact position at which the cur-
rently displayed vessel image was acquired. To provide visual feedback regarding the
current C-arm/image alignment, an “alignment signal” (Fig 1 (b)), shaped like an ar-
row head pointing to the left is used. It is divided in two parts: The upper half repre-
sents the alignment with respect to the Left/Right image plane, while the lower half
represents the alignment with respect to the Head/Feet image plane. The degree of
alignment is expressed in a color code as well as in an exact angle value. A dark blue
and light blue color represents a “very good” (≤ 0.5˚) and “good” (≤ 1.0˚) alignment,
while the color red signals an “unacceptable” (> 1.0˚) alignment. Roadmapping is
640 M. Kukuk and S. Napel
only possible for a “very good” or “good” alignment. We implemented two solutions
to the problem of fast and accurate C-arm alignment that appear to have their respec-
tive advantages and shortcomings and are therefore in practice commonly used inter-
changeably: automatic and interactive alignment.
Automatic alignment is performed in two steps, corresponding to RoRo Phases 1
and 2. First, the interventionalist uses the RoRo joystick to browse through the avail-
able projections in order to find the most suitable working view. Then, he/she clicks
on the “2 - Move-C-arm” button in the control panel (Fig. 1 (c)). This sets the C-arm
system into “automatic run” mode. Simply deflecting the C-arm joystick will auto-
matically move and stop the C-arm at the correct position. After the position has been
reached, the “Alignment signal” switches to “dark blue” and roadmapping can begin.
Interactive alignment is done in only one step. As the interventionalist moves the
C-arm back and forth along the acquisition plane, the image closest to the current C-
arm position is displayed. This is perceived as the vessel tree “following” the C-arm.
In other words, browsing and aligning are done in parallel. As can be seen in Fig 2,
top right, if the difference d between two images is sufficiently small (e.g 1.5˚),
roadmapping will be possible for any C-arm position within acquisition range. If d is
greater than 2.0˚, there will be C-arm positions, for which the alignment is “unaccept-
able” (red signal) and therefore roadmapping will not be allowed. In this case, the in-
terventionalist interactively fine-tunes the C-arm position using the alignment signal,
until a “very good” or “good” alignment is reached.
While automatic alignment is exact, it requires the use of two different joysticks
and is therefore relatively slow. On the other hand, interactive alignment only requires
the use of the C-arm joystick but has an associated “learning curve”. In most cases,
images can be acquired at small enough intervals (d ≤ 1.5˚), in which case no fine-
tuning is required and alignment is achieved once the working view has been found.
Although visualization is limited to 2D image display, several challenges for a variety
of display modes need to be addressed. For a smooth display of all guidewire manipu-
lations, fluoroscopy frame rates of 30fps are often used. To match this frame rate,
OpenGL texture mapping was employed for image display. To additionally allow for
the display of non-square, non-power-of-two projections (limitation to OpenGL ver-
sions before 2.0), images are broken up into square, power-of-two tiles, which are
then individually mapped to the screen. During all display modes, basic display func-
tionality, such as window/level, zoom and pan is available. It is of particular clinical
importance to allow the interventionalist to change the imaging system’s zoom format
or SID (source image distance) after the acquisition of the rotational run. RoRo then
automatically reformats the DSA projections in real time, to reflect the changes.
Additionally, a stereo visualization mode allows the interventionalist to perceive
vessel trajectories in 3D. This is achieved by simply selecting two projections as a left
and right eye stereo-pair and displaying them such that each eye only sees its respec-
tive image. We implemented a stereo projection system using polarizing filters in
front of two DLP projectors with matching polarized glasses and the red/blue (ana-
glyph) technique, using red/blue glasses.