New sophisticated tools are increasingly developed for diagnostic and therapeutic percutaneous interventions, with a wide range of applications, such as biopsy and thermal ablation [1
]. Navigation software and robotic assistance may enhance puncture accuracy, increase patient safety and reduce radiation exposure as well as procedure duration [5
]. Precise planning and accurate needle placement are key for ablative procedures when multiple needles need to be placed at once in order to achieve overlapping ablation zones. By increasing the speed and accuracy of multiple placements, large tumors and multifocal diseases can be treated efficiently [6
Image-guided needle placement may be based on CT, MRI, ultrasound, and X-ray. MRI provides an excellent soft tissue resolution but it is expensive and requires MRI compatible instruments [7
]. In ultrasound, it is not always possible to visualize the tumor adequately due to shadowing artefacts caused by air or bone. Many different interventional procedures are routinely performed in a CT environment, delivering three-dimensional (3D) contrast-enhanced high quality image data. Navigation systems may assist in planning and needle placement, and help to reduce the radiation exposure to clinical staff and patients [8
A variety of guidance and navigation systems have been developed using active systems with visual or electromagnetic tracking. The robotic approach offers several advantages over other assistance systems. No direct line of sight is needed and the ability to function is unaffected by the presence of ferrous materials [9
]. Other passive, patient-mounted needle holders have been developed [10
] to improve interventional accuracy taking into consideration a simplified procedure with lower acquisition costs.
The aim of this study is to compare the accuracy of a commercially available robotic system (Maxio, Perfint Healthcare, Chennai, India) with a low-cost patient-mounted system for CT-guided needle navigation (Puncture Cube System, Medical Templates AG, Egg, Zurich, Switzerland). In addition, the usability, handling time and drawbacks of both techniques are discussed.
In total 560 punctures were performed: 320 punctures were assessed using the Maxio robotic system, including 80 punctures for 1.25, 2.5, 3.75 and 5 mm slice thickness, respectively; 240 needles were placed using the PCS including 80 punctures for 1, 3 and 5 mm slice thickness, respectively. Results are expressed as mean ± SD.
The achieved accuracies measured with the normal and Euclidean distance are summarized in Table 1
The Kolmogorov–Smirnov Test showed no normal distribution. With histograms, this could be verified, showing mostly right-skewed data for the ED and ND with some outliers.
Using the Mann–Whitney U test, no significant differences between accuracy were observed, comparing all slice thicknesses.
ED: 1.25 mm vs. 2.5 mm, p = 0.56; 1.25 mm vs. 3.75 mm, p = 0.71; 1.25 mm vs. 5 mm, p = 0.44.
ND: 1.25 mm vs. 2.5 mm, p = 0.57; 1.25 mm vs. 3.75 mm, p = 0.81; 1.25 mm vs. 5 mm, p = 0.79.
The achieved accuracies measured with the normal and Euclidean distance are summarized in Table 2
The Kolmogorov–Smirnov test showed normal distribution. Using Student’s t-test, no significant differences between accuracy were observed, comparing all slice thicknesses.
ED: 1 mm vs. 3 mm, p = 0.13; 1 mm vs. 5 mm, p = 0.11; 3 mm vs. 5 mm, p = 0.98.
ND: 1 mm vs. 3 mm, p = 0.08; 1 mm vs. 5 mm, p = 0.06; 3 mm vs. 5 mm, p = 0.99.
3.3. Maxio vs. PCS
Comparing both systems using the Mann–Whitney U test, we observed a significant difference between accuracy for all slice thicknesses (Figure 4
ED (Maxio vs. PCS): 1.25 mm vs. 1 mm, p < 0.001; 2.5 mm vs. 3 mm, p < 0.001; 3.75 mm vs. 3 mm, p < 0.001; 5 mm vs. 5 mm; p <0.001.
ND (Maxio vs. PCS): 1.25 mm vs. 1 mm, p < 0.001; 2.5 mm vs. 3 mm, p < 0.001; 3.75 mm vs. 3 mm, p < 0.001; 5 mm vs. 5 mm, p < 0.001.
The PCS punctures required about 30% less time (mean 158.63 s [SD ± 23.38] vs. 225.67 s [SD ± 17.2]). Additional time is needed to physically move and dock the robotic device to the CT tableside, taking approximately 3–4 min [15
]. It required approximately the same amount of time to switch off the device, undock and move it to the parking location.
In total, all 560 punctures were performed successfully. Our results of a phantom puncture series show that the Maxio yields greater accuracy compared with the PCS and other image-guided intervention systems. On the other hand, punctures with the PCS required less time.
The ND might be the most reliable indicator of accuracy. In fact, the aluminum tips prevent the needle from moving forward once they are accurately placed. On the other hand, when not touching the cone, the needle could tend to slip slightly through the gelatin-filled phantom, while the phantom is moved. Fortunately, this did not occur in our experimental setup but could be the case in a millimeter/submillimeter range, even if the needles were embedded tightly in a rigid gel environment.
It is arguably difficult to compare the results of our two systems with those of existing studies, as different setups were used, and different endpoints were evaluated.
Our group used the identical phantom with different navigation systems (Table 3
). As far as technically possible, we applied the same method to ensure comparability. As an example, Stoffner et al. [11
] used a robotic assistance system, the Innomotion, showing accurate results comparable to the Maxio. Moreover, Putzer et al. [13
] investigated the accuracy of two electromagnetic navigation systems, the AxiEM and PercuNav, with comparable results to the less complex PCS.
Overall, the Maxio turned out to be the most accurate guidance system tested by our group.
Looking at the data for the PCS, there is no significant difference between accuracy, comparing all three slice thicknesses. The difference of the mean value does not exceed 0.5 mm for the ED and 0.6 mm for the ND.
Mokry et al. [16
] came to the result that the PCS improves accuracy and reduces intervention time compared to the free-hand method. Their phantom study showed almost equal results compared with ours (3.4 mm ± 2.3 mm vs. 3.84 mm ± 1.75 mm).
Our results show that the precision to arrive at targets decreases with greater target depth. In fact, the modular construction and the plastic material have a negative impact on its stiffness, increasingly affecting accuracy with longer needle paths.
In contrast to the Maxio, the precision to arrive at the target with the PCS depends on its depth due to a limitation of adjacent valid combinations with the templates. For a chosen depth of 10 cm, the distribution of errors can be color coded graphically (Figure 5
), showing a decrease in valid combination in the periphery.
In a clinical routine, it is important to attach the PC to the patient skin with approximate prior knowledge of the target. The small size of the PCS and its limited range due to a prescribed number of combinations make it crucial to find the right spot. Bony landmarks, prior examinations or topogram images can help to position the device. On the other hand, uneven anatomic locations can cause difficulties attaching the PC due to its flat template. Once the PC is mounted, patient movement will have less impact compared with the Maxio.
The PCS is faster to use. Taking into account the time the Maxio needs to dock and undock, the PCS allows needle placement considerably faster. Furthermore, the PCS allows the operator to plan and position the probes simultaneously when two physicians are present, making it possible to gain speed. The robotic positioning process of the Maxio is instead fully automatic and does not depend on the experience of the physician.
Looking at the data for the Maxio, no significant difference can be observed when comparing all four different slice thicknesses. The mean ED and ND show highly accurate results from 1.25 to 5 mm slice thickness. Comparing all slice thicknesses, the difference of the mean value does not exceed 0.25 mm for the ED and 0.1 mm for the ND.
A cadaver study was performed by Croissant et al. [17
], revealing highly accurate results using the Maxio equipment during spinal interventions. The robot-assisted placement of 24 K-wires showed a mean deviation of 1.2 mm in the horizontal-axis and a mean deviation of 0.5 mm in the vertical-axis, comparable with our findings.
Comparatively, the robotic device did not show any disadvantages in accuracy for a higher insertion depth or a particular angulation. The robotic arm is capable of moving freely within 5 degrees of freedom. It has a large range of achievable orbital needle angles, ranging from −95° to 95° in both lateral and craniocaudal directions. In comparison, the grid of the PCS limits oblique needle insertion to approximately 45°. The Maxio is capable of moving its arm independently, parallel to the CT cradle with a range of 180 mm and perpendicularly 300 mm from the gantry center line to the opposite side of the docked device.
The software of the Maxio is limited to planning up to six needles simultaneously, which is a disadvantage for ablation of larger tumors or multifocal disease. Larger, more perfused tumors need multi-probe needle placement. A single-pass needle insertion is not the standard procedure for complex clinical needle placements, such as stereotactic radiofrequency ablation of liver tumors.
Once docked, the Maxio limits physical access on the mounted side of the CT table. This downside is similar to other guidance devices, although the Maxio is quite large, with dimensions of 85 cm × 80 cm × 180 cm, when docked. The quite unwieldy end-effector of the Maxio makes it impossible to plan multiple needle insertions in a small entrance area due to collision. On the other hand, the PCS allows the placement of a considerably higher number of probes simultaneously yet is limited by the small design of the template and the number of holes.
The Maxio allows the operator to manually correct all six needle positions after placement. The PCS also offers the possibility of adjusting the position of the needle as the upper template can be brought in contact to the lower plate. However, this is not possible when more than one needle is obliquely embedded inside the template.
The planning process is simple on both systems and can be operated by one person. The workstation of the Maxio allows the operator to define the target and entry points, which subsequently executes the planned trajectory, without requiring a physician’s calculation of entry-to-target distance and angulation. In comparison, the PCS only allows single-probe access per treatment plan.
Compared to conventional puncture techniques and low-cost systems, robotic systems have much higher capital costs. The acquisition costs for the PCS are considerably lower compared to the Maxio.
Radiation exposure to the physician can be virtually eliminated with both systems. Conventional needle placement using CT fluoroscopy or incremental needle advancement, needs multiple scans for each probe. Both systems allow several probes to be placed simultaneously. Only a planning scan and a verification scan are generally required, reducing radiation exposure to the patient. Furthermore, the systems yield comparable accuracy using bigger slice thickness, further reducing radiation exposure. On the other hand, planning on a 5 mm CT data set turned out to be more time consuming. The lower spatial resolution made it harder to locate our target points. All three planes had to be examined to estimate the exact spots.
Some limitations are worth mentioning due to the ex vivo environment. A limitation of this phantom study is the inability to consider respiratory movement. Input errors, due to respiratory motion, are key points in the clinical setting. Different techniques beyond active or passive movement could not be applied during this phantom study. Gating the patients’ breathing is a key component of normal clinical routine.
Unlike the flat cover of the phantom, uneven human body surfaces could cause difficulties attaching the PCS. In a clinical routine, inaccuracies might even increase due to the movement of the skin when placed on a patient. Gelatin, which is the most commonly-used phantom material in the literature, shows similar behavior to human soft tissue. Nevertheless, the living body consists of many kinds of tissue components. Therefore, tissue may be movable especially at the interface of the tissues, which may cause needle bending and target movement during needle insertion.