Integrated IVUS catheter with EM sensor and image acquisition system

To measure the 3D position and orientation of the IVUS catheter, a six degrees of freedom (6DOF) EM sensor (0.8 × 9 mm, NDI Corporation, Canada) was attached to the outer surface of the IVUS catheter with an EM tracking system (Aurora Tracking System, NDI Corporation, Canada) (Fig. 1a). Images were acquired using a 40 MHz IVUS catheter (Opticross, Boston Scientific, USA) with a 2.6 Fr imaging window profile. The video stream was transmitted from the IVUS system to a computer using a video capture card (DVI2USB, Epiphan Systems Inc., Canada).

Fig. 1

a Experimental setup for the evaluation of EM sensor interference caused by IVUS to the EM sensor; b Integrated catheter in two cases. The tape used for fixing is not shown; c Wire showing an artifact on an IVUS image as a dot. The contrast has been adjusted

Evaluation of interference from a mechanical IVUS to EM tracking system

To quantify the amount of interference from IVUS to EM tracking system, specifically EM and mechanical (due to vibration) interference, we placed the EM sensor on a test bench along with the EM field generator of the tracking system. The distance between the EM sensor and field generator was 75 mm (Fig. 1a). The position of the field generator was adjusted such that it was on the cross section passing through the center of the catheter. The position, rotation, and indicator of magnetic field disturbances were recorded using the EM tracking system software (NDI Toolbox 5.004.015) in 10 s by changing the position of the IVUS transducer relative to the EM sensor. The indicator was defined in the software as the magnitude of interference caused by IVUS. The outputs were recorded under two different conditions:

(1)

Nonoperating IVUS

To evaluate the effect of the metallic components of the IVUS transducer and basal operation current of the entire IVUS system, the IVUS was not performed while the main power switch of the IVUS system was turned on.

(2)

Operating IVUS

To evaluate the interference due to the driving current of the IVUS transducer unit and its rotating motion, the IVUS transducer was rotated, and image acquisition was conducted.

Preparation of two possible arrangements of the integrated catheter

The mechanical IVUS image and EM sensor mounted on the outer IVUS catheter were not fixed in a rigid frame during pullback. The rotation of the image coordinate system relative to the outer IVUS catheter should be obtained in real time to register the IVUS images to the EM tracking coordinate system (fixed on the field generator). A fiducial marker must be integrated to determine the relative rotation. To minimize the diameter of the integrated catheter, two possible arrangements of the position for the EM sensor and IVUS transducer are proposed in Fig. 1b:

Case 1: The IVUS transducer is located on the distal side just in front of the EM sensor. An additional wire marker is integrated on the distal side.

Case 2: The IVUS transducer is located on the proximal side just behind the EM sensor. The electrical cable for power and signal transmission of the EM sensor is used as a fiducial marker.

These two possible designs were evaluated in terms of the reconstructed surface of a rigid vascular phantom. For Case 1, a silk wire was fixed approximately 0.3 mm above the IVUS transducer as the fiducial marker.

Registration of IVUS images to the EM tracking coordinate system

The fiducial marker (silk wire for Case 1, electrical cable for Case 2) created an artifact, a dot in this case, on the IVUS image. The position of the dot was marked manually in the first frame, and could be tracked by finding the center of the area with a high intensity near it in the previous frame. The distance from the dot to the origin of the image was fixed such that the position only needed to be searched in the circumferential direction.

The origin of the IVUS transducer was regarded at the center of the black circle in the IVUS image, which could be detected using the Hough transform [7]. We connected the dot to the origin and measured the artifact angle from the x-axis to the connection line as \(_^\) (Fig. 1c). To obtain the relative rotation angle \(_^\) from the x-axis in the EM sensor coordinate system to the x-axis in the image coordinate system, the origin of the image was translated to the origin of the EM sensor (\(_\) in Fig. 1b) such that the angle \(_^\) is.

$$_^=\pi -(_^+_^)+_^$$

(1)

where \(_^\) and \(_^\) are constant angles from the artifact and EM sensor to the connection of the IVUS and EM sensor counterclockwise, respectively.

As the normal vector of the image is parallel to the z-axis of the magnetic sensor, the rotation matrix of \(_^R\) is a function of the rotation angle \(_^\) along its axis.

$$_^R=\left(\begin\text_^&\quad -\text_^&\quad 0\\ \text_^& \quad \text_^& \quad 0\\ 0& 0& 1\end\right)$$

(2)

Let \(_^t\) be the displacement from the origin of the image coordinate system to the EM sensor.

$$_^t=_}\left(_^\right),_}\left(_^\right),_\right)}^$$

(3)

\(_^t\) can be obtained from the measurements of the radial distance \(_\) and the axial distance \(_\) between the origin of the IVUS catheter and EM sensor. These parameters can be calibrated by registration with a model of a known size (refer to the next section). Let the coordinates of one point on the lumen contour in the IVUS image coordinate system be \(_\) and the coordinates of the corresponding point in the EM tracking coordinate system be \(_\). The transformation from \(_\) to \(_\) is given by

$$_=_^R \cdot \left(_^R\cdot _+_^t\right)+_^t$$

(4)

where \(_^R\) and \(_^t\) are the rotation and translation of the output of the 6DOF EM sensor, respectively.

Identification of geometrical parameter

According to Eqs. 1–4, four constant geometrical parameters, namely \(_^\), \(_^\), \(_\), and \(_\) should be identified for the integrated catheter to determine the transformation from the image coordinate system to the EM tracking coordinate system. \(_\) can be determined by physical measurements. As the origin of the sensor is not located at its center and may be different for each sensor, \(_\) cannot be directly measured. These three parameters can be calibrated using a calibration object of known size [10], which has the same shape as the vascular phantom (Fig. 2), but with all side branches removed. This process includes the following steps.

1.

Estimate the range of parameters based on the geometry of the integrated catheter.

2.

Pullback the integrated catheter inside the phantom (the distance to the field generator was 75 mm) with a random twist and make a reconstruction.

3.

Register the reconstructed point cloud \(_}}\) with initial values to the dense point cloud \(_}}\) from the geometrical data of the phantom using the iterative closest point (ICP) algorithm [15] and obtain the rotation \(}\) and translation \(t\).

4.

Calculate the mean square error (MSE) of all the vertices on the reconstruction.

$$\begin }& =\underset},t}}}\frac\sum\Vert _}}\\ & \qquad -(}\ \cdot _}}(_^\text _^\text_)+t)\Vert^\end $$

(5)

5.

Use an optimization algorithm to determine the minimum MSE for the optimized parameters.

Fig. 2

Vascular phantom and processing method for 3D intravascular reconstruction

Three-dimensional surface reconstruction of a vascular phantom

A rigid phantom made of transparent resin was placed at 75, 125, and 175 mm parallel to the field generator. It was a simulated model artificially designed from a carotid artery with multiple side branches (Fig. 2). Its cross section was elliptical, with different major and minor axes in each part of the vessel (diameter range: 3–7 mm). The eccentricity was always greater than 0.8.

For each IVUS image frame, the artifact of the marker was first segmented. Subsequently, the artifact was erased from the image to segment the inner lumen contour using a method similar to the one in [7]. All the contours contained the same number of points. The distance from the center of the frame to each point should be adjusted according to the liquid and temperature [16]. The relative rotation of the wire to the image coordinate system \(_^\) was also measured from the image. Subsequently, each frame was assigned the position and orientation of the EM sensor at that moment, which enabled the calculation of the contour points in the EM tracking coordinate system.

The entire pullback process was performed manually. The pullback speed of the IVUS catheter with the EM sensor was not constant. Stick slip and reversed directional motion were observed in the experiments. By selecting the appropriate contours and connecting them, a 3D vessel model could be correctly reconstructed in real time. After each pullback, the ICP algorithm [15] was applied to register the reconstructed model with the phantom geometry, as shown in Eq. (5). The reconstruction error was evaluated by the signed distance \(d\) from each vertex on the reconstructed model after registration to its closest point on the mesh surface of phantom geometry \(_}}\).

$$d=\sigma \Vert _}}-(}}^}\ \cdot _}}+^)\Vert $$

(6)

Here, when the vertex is outside the phantom geometry, \(\sigma =1\); when the vertex is inside the phantom geometry, \(\sigma =-1\). \(_}}\) is the vertex on the reconstructed model before registration. \(}}^}\) and \(^\) are the rotation matrix and translation vector, respectively, calculated by registration.