Probe versus microscope: a comparison of different methods for image-to-patient registration

Probe versus microscope: a comparison of different methods for image-to-patient registration Purpose Computer-aided navigation is widely used in ENT surgery. The position of a surgical instrument is shown in the CT/MR images of the patient and can thus be a good support for the surgeon. The accuracy is highly dependent on the registration done prior to surgery. A microscope and a probe can both be used for registration and navigation, depending on the surgical intervention. A navigation system typically only reports the fiducial registration error after paired-point registration. However, the target registration error (TRE)—a measurement for the accuracy in the surgical area—is much more relevant. The aim of this work was to compare the performance of a microscope relative to a conventional probe-based approach with different registration methods. Methods In this study, optical tracking was used to register a plastic skull to its preoperative CT images with paired-point registration. Anatomical landmarks and skin-affixed markers were used as fiducials and targets. With both microscope and probe, four different registration methods were evaluated based on their TREs at 10 targets. For half of the experiments, a surface registration and/or external fiducials were used additionally to paired-point registration to study their influence to accuracy. Results Overall, probe registration leads to a smaller TRE (1.69 ± 0.74 mm) than registration with a microscope (2.19 ± 0.94 mm). Additional surface registration does not result in better accuracy of navigation for microscope and probe. The lowest mean TRE for both pointers can be achieved with paired-point registration only and radiolucent markers. Conclusion Our experiments showed that a probe used for registration and navigation achieves lower TREs compared using a microscope. Neither additional surface registration nor additional fiducials on an external reference element are necessary for improved accuracy of navigated ENT surgery on a plastic skull. Keywords Registration · Microscope · Navigation · Target registration error Introduction skull base to support the surgeon with detailed images of the surgical area. In this case, the pointer can be the microscope Clinical navigation systems are commonly used for ENT or a probe. interventions. At critical regions, they are supporting the Prior to using a navigation system, the CT/MR images surgeon by realizing the exact positions of a pointer inside have to be registered to the patient. In general, the two main the patients’ body on the CT/MR images. Depending on the registration methods are paired-point and surface registration surgical intervention, different pointers can be used for nav- [1,2]. For the first method, points on the patient (‘fiducials’) igation; for example, a microscope is required at the lateral have to be located, together with the corresponding points in the images. Thus, the rigid transformation between image and patient coordinates can be calculated. Besides, anatom- B Martina Perwög martinaperwoeg@hotmail.com ical structures also skin-affixed or bone-implanted markers can be used as fiducials. Skin-affixed markers are known to ENT/4D-Visualization, Medical University Innsbruck, be prone to changes of the skin according to different hydra- Anichstr. 35, Innsbruck, Austria tion statuses or effects of muscle relaxing agents and thus are ENT, Klinikum der Universität München, Munich, Germany to be used with caution [3]. Those aspects can be avoided if ENT, Kardinal Schwarzenberg Klinikum, Schwarzach im bone-implanted fiducials are used, but because of the inva- Pongau, Austria 123 1540 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 Table 1 Overview of the different experiments with the number of fiducials and targets used for registrations and measurements Pointer # Fiducials # Targets Probe Microscope Anat. Donut Radiolucent Surface reg. VBH Anat. Donut Radiolucent A1 B1 5 2 3 – – 5 3 2 A2 B2 5 2 3 25 – A3 B3 3 1 2 – 4 A4 B4 3 1 2 25 4 siveness they are not the first choice, due to side effects for the patient. The most commonly used fiducials are anatomi- cal landmarks, as they are easy to use and non-invasive. Surface registration helps automation and registers a larger number of points to the image. These points are collected by moving a pointer over a large surface on the patient, like the forehead. With the iterative-closest-point (ICP) algorithm, the surface of the patient is matched to the corresponding 3D-image-model [2]. For interventions, e.g. the lateral skull base or the petrous bone, the VBH (Vogele–Bale–Hohner) head holder can be used additionally to fix the patients head on the OR Table [4]. External fiducials can be affixed to the VBH mouthpiece to get a larger area of fiducials that surrounds the surgical Fig. 1 Experimental set-up for experiments A1,…, A4. The plastic region and thus improve accuracy [5,6]. More details are skull is fixed with the VBH head holder on the OR table. The camera given in the ‘Materials and methods’ section. is optimally placed and the navigation system placed opposite to the The ideal case for the surgeon would of course be per- surgeon [23] fect (error free) navigation, but the registration is a critical step and has to be done as accurately as possible. However, Two different pointers were used for both registration and it always introduces errors as calibration, tracking and mea- TRE measurement: a microscope and a pointer. Both pointers surement errors can occur, but user localization errors are are optically tracked by a camera relative to a DRF (digital also common in surgical navigation [7–10]. reference frame), so the coordinates of the points in patient After registration, the typical navigation systems report space can be determined and transformed to image space. a single error which is called the fiducial registration error Four different registration methods have been experimen- (FRE) [11]. The FRE is the RMS distance of CT image fidu- tally evaluated with each pointer (probe or microscope), two cials and patient fiducials that are transformed to the CT of them with external fiducials on the VBH mouthpiece. image. However, the target registration error (TRE) is much Surface registration was additionally used for half of the more important for the surgeon in practice. It reports the experiments (see Table 1 for a detailed description). The TRE accuracy at a target location close to the surgically relevant was measured on 10 different targets, and all methods were positions and is uncorrelated with the FRE [12]. compared to each other. Thus, it was possible to evaluate the Unfortunately, not many studies regarding the applica- performance and differences of the registration methods and tion accuracy (the TRE) of different registration methods pointers used for our experiments. have been done comparing microscope and probe registra- tion methods [13–19]. Therefore, the aim of this study was to find the optimal Materials and methods registration method of a plastic skull to its preoperative CT images in terms of application accuracy at the lateral skull The experimental set-up can be seen in Figs. 1, 2. base, concerning the pointer and the registration method. Two different questions were evaluated: What kind of pointer Phantom should be used for registration and measurements? Which registration method is the optimal method to register a plas- A plastic skull (Somso, Coburg, Germany) was used for all tic skull to its CT images? experiments. Radiolucent markers (X-spot, izi-Kooperation, 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1541 Fig. 3 Plastic skull with the VBH mouthpiece. Targets are shown in pink; fiducials are shown in black. The number in green shows the Fig. 2 Experimental set-up when the microscope is used for the regis- fiducial used for registration only with VBH mouthpiece. (fiducials: tration. The DRF is fixed on the left side of the microscope (the blue 3—X-spot on the frontal bone, 4—X-spot on the left nasal bone, 6— cylinder hidden in this figure on the left side) [23] anatomical landmark on the right supraorbital foramen, 7—right inner cantus, 8—anterior nasal spine, 9—saddle point on the left frontozy- gomatic suture, 10—infraorbital foramen; targets: 1, 2—radiolucent markers on the right frontal bone and the right maxillary bone, 3— Beekley Corp., Conn, USA) and MR marker (Donuts, izi- donut marker on the left frontal bone, 6—left supraorbital foramen, Kooperation, Beekley Corp., Conn., USA) were attached on 7—left inner canthus, 8—left nasal bone, 9—saddle point right fron- tozygomatic suture, 10—right infraorbital foramen) [23] the skull. For all experiments the plastic skull was fixed on an OR table with a VBH head holder [4]. With the VBH mouth- piece, the skull can be attached at the holder [6]. Dental foam Navigation system vacuum fixation on the maxilla leads to a stable construction in all experiments. This is a fast and secure non-invasive The SNN navigation system (Fa. SNS, Aalen, Germany) was method to fix a patient during interventions. On the refer- used for navigation with optical tracking of (flashpoint-3000- ence element of the VBH mouthpiece, radiolucent markers system, IGT, Boulder, USA). It comprises classical pointer- were attached (see Figs. 3, 4). Those external markers were based navigation and a navigated microscope with heads-up only used for half of the experiments (see Table 1). display capabilities for navigation. Pointers CT images Two different pointers were used for registration and naviga- CT images of the plastic skull were acquired with a Siemens- tion on a plastic skull. The probe (135-mm flashpoint) can be Sensation-16 (Siemens, Erlangen, Germany), leading to 2 optically tracked by the camera with the two LEDs attached axial datasets. The first dataset was captured with the refer- to it and is connected with a cable to the navigation system. ence element fixed on the VBH mouthpiece in place and the The microscope (OPMI 2000, Zeiss, Oberkochen, Ger- second one without the reference element. Image parameters many) can be optically tracked with a DRF attached to it (see were 140 kV, 220 mAs with 1 mm slice thickness. Fig. 2). The microscope is factory-calibrated, and thus, the 123 1542 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 and the microscope was not moving anymore, this point was used for registration. In detail (cf. Figs. 3, 4), the fiducials used for the experi- ments without VBH mouthpiece are the 2 donut markers on the right parietal bone and the right frontal bone, the 3 radi- olucent markers on the frontal bone, the left nasal bone and the left temporal bone. The five anatomical landmarks are the right supraorbital foramen, the right medial cantus, the anterior nasal spine, the ‘saddle point’ on the left frontozy- gomatic suture, and the left infraorbital foramen. For the experiments with the VBH mouthpiece, the fidu- cials in detail are 4 radiolucent markers on the referencing element of the VBH mouthpiece, 1 donut marker at the right parietal bone, 2 radiolucent markers median on the frontal Fig. 4 Plastic skull from the right with fiducials (black) and fiducials on bone and the left temporal bone and the 3 anatomical land- the VBH mouthpiece (green). (1—Donut marker on the right parietal marks on the anterior nasal spine, the left infraorbital foramen bone, 2—Donut marker on the right frontal bone, 3 and 4—X-spots on the VBH—mouthpiece used only for registration with the VBH and the right supraorbital foramen. mouthpiece) [23] For the surface registration, 25 points were acquired by the probe touching the mid-face, the frontal bone and the orbitae of the skull surface. The same anatomical locations were parameters of the camera are known and the position of the focused with the microscope with maximum magnification. point that is in focus can be determined. Again 25 points were selected, when the points were in focus and the microscope was not moving. Registration TRE measurements To register the plastic skull to its CT images, four different After all different registration methods, the TRE was mea- registration methods were performed with the probe (group sured on 10 targets, which were per definition different to the A) or the microscope (group B). As a skull has only a few fiducials (c.f. Fig. 3). The targets used were always the same prominent anatomical landmarks that can be defined well for A1–A4 and B1–B4. both on the patient and in the CT images, different kinds Five anatomical landmarks, 2 radiolucent markers and 3 of markers were used as fiducials (but also targets, as can donut markers were used. In detail, the targets (in ascending be seen in the next section), to have a sufficient amount of order target number 1–10) are the 2 radiolucent markers on fiducials for each registration method. the right frontal bone and the right cheek bone, the donut The different methods and fiducials used for registration markers on the left frontal bone, near the left coronal suture were defined as follows (see also Table 1): and the left parietal bone. The 5 anatomical targets are the A1 and B1: paired-point registration with 5 anatomical left supraorbital foramen, the left medial cantus, the bony landmarks, 2 donut markers and 3 radiolucent markers structure on the left nasal bone, the ‘saddle point’ on the right A2 and B2: paired-point registration with 5 anatomical frontozygomatic suture and the right infraorbital foramen. landmarks, 2 donut markers, 3 radiolucent markers and addi- Each experiment (registration and TRE measurement) was tionally a surface registration with 25 points repeated 10 times. To avoid a learning effect during the A3 and B3: paired-point registration with 3 anatomical experiments, initial experiments were done to get used to landmarks, 1 donut marker and 2 radiolucent markers on the the equipment without recording the results. skull and 4 radiolucent markers on the VBH mouthpiece Each target was located as exactly as possible with the A4 and B4: paired-point registration with 3 anatomical probe. A screenshot of the navigation systems display was landmarks, 1 donut marker and 2 radiolucent markers on the taken, when the probe was placed at the target. The same skull, 4 radiolucent markers on the VBH mouthpiece and procedure was repeated with the microscope, and each tar- additionally a surface registration with 25 points get was located with the maximum magnification and with The probe and the microscope were both used as point- autofocus. ers to locate a fiducial on the plastic skull. The fiducials After finishing the experiments, each screenshot was pro- were physically located with the tip of the probe. With the cessed in Adobe Photoshop 5.0. Each image plane was microscope, the fiducials were located with maximum mag- maximally magnified. The difference of real target and pro- nification. Only when the fiducial of interest was in autofocus jected target (Δx, Δy, Δz) was measured in each image 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1543 Fig. 5 Screenshot of the navigation system’s monitor when pointing on the X-spot. The difference between the target point and the pointer tip is measured in axial, coronal and sagittal views (c.f. Fig. 6, right). The crosshair is used to get the scaling for the measurement (c.f. Fig. 6,left) [23] 2 2 2 plane, axial (x–y), sagittal (x–z) and coronal (y–z). So each TRE(t ) = x + y + z , ij ij ij ij x-, y- and z direction was measured twice (10 × 6 data matrix for each group). Those pixel measurements were converted which is the TRE for target number i, i = 1,..., 10, experi- in mm. The scale is known by measuring the length of the ment j, j = A1,..., A4, B1,..., B4. crosshair, which is 1 cm. An example of the TRE measure- All calculations were done with all measurements of each ment is shown in Figs. 5, 6. group in Matlab. It could be possible that the target was not visible in one Besides that, a worst-case scenario was analysed, where plane, and thus, the error could only be measured in the other only the largest values in each x-, y-, and z-directions of each two planes. The missing values were assigned as NaN (not-a- repetition were used for calculations (10 × 3 data matrix for number) and replaced with the corresponding value in x-, y- each group). and z-directions; for example, if the axial x-direction could not be measured, the measurement of the sagittal x- direction Evaluation was used for both x-values. Overall 4800 measurements were taken, 600 for each The aim of this investigation was to find the difference regard- experiment. For each target, the mean of the 10 repetitions in ing the TRE between registrations via microscope and probe each direction was calculated. Thus, the values (x , y , ij ij on a plastic skull. Besides that, it was evaluated if surface z ) remained for each target. With these values, the TRE ij registration (additionally to paired-point registration) and was calculated by additional fiducials fixed on the reference element of the 123 1544 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 Fig. 6 Left side: measuring the length of the crosshair, which is defined the navigation system. The probe (blue) is pointing at the target (white as 1 cm. The measured value is 1.62 in this image and is used as the elliptical point). The difference between the tip of the probe and the scaling factor for the measurements. Right side: example of the TRE centre of the target is 0.05 [23] measurement on a X-spot. Maximized part of the image of the display of Table 2 TREs with the p values of the Wilcoxon signed-rank test TRE (mm) A1 B1 A2 B2 A3 B3 A4 B4 Mean ± std 1.41 ± 0.61 1.70 ± 0.68 1.48 ± 0.61 3.02 ± 1.37 1.44 ± 0.50 1.92 ± 0.72 2.44 ± 1.09 2.10 ± 0.81 p value 0. 0645 0.0020 0. 0273 0.1602 It can be seen that A2–B2 and A3–B3 are statistically not equal. (H :A = B , i = 1,..., 4). Mean values and standard deviations are given in mm 0 i i VBH mouthpiece improve the TRE. Group A (probe) and ference of microscope- and probe-based registration can also group B (microscope) were compared to each other. The over- be observed (Wilcoxon signed rank, p < 0.0001). all TRE over all four types of registration was calculated for a The overall TRE for group A is 1.69 ± 0.74 mm, and for first overview, but also the TRE of each registration method group B, it is 2.19 ± 0.94 mm. was calculated separately. All calculations were done with The mean TRE of A1 is 1.41 mm compared to B1 with Matlab R2012a. Statistical tests were done with α = 0.05. a mean TRE of 1.70 mm. The mean TRE of A2 is 1.48 mm An ANOVA was used to find differences in group A or group compared to B2 with a mean TRE of 3.02. A3 and B3 have B. If a statistically significant difference was found, multiple a mean TRE of 1.44 mm and 1.92 mm, respectively. A4 and comparisons were done with Bonferroni correction to anal- B4 have a mean TRE of 2.44 mm and 2.10 mm, respectively yse which experiments were significantly different to each (Table 2). other. For comparing Ai and Bi, i = 1,..., 4, a Wilcoxon These results are leading to statistical results as follows: signed-rank test was used. A1 is equal to B1 ( p = 0.0645) and A4 is equal to B4 ( p = 0.1602). A2 and B2 are significantly different ( p = 0.0020), as well as A3 and B3 ( p = 0.0273). The four probe groups (A1–A4) are significantly different Results to each other ( p = 0.0004). A4 has the largest error. B1–B4 are significantly different to each other ( p = All results are mean values with their standard deviation and 0.0012), and B2 has the largest TRE. can be seen in detail in Tables 2, 3. Group A defines the Regarding the error in x-, y- and z-directions, it could be registration with the probe and group B the registration with observed that overall the TRE is anisotropic. For the probe the microscope. measurements, the error in y-direction is the smallest with All errors in the groups, except A1, A2 and B2, are nor- 0.79 ± 0.49 mm and in x- and z-directions the error is 1.00 ± mally distributed ( p < 0.05). All 8 groups compared to each 0.56 mm and 1.03 ± 0.62 mm, respectively. This is also valid other are not equal (ANOVA, p < 0.0001); an overall dif- 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1545 Table 3 Mean TRE at different landmarks used as targets for the 8 groups TRE (mm) A1 B1 A2 B2 A3 B3 A4 B4 Radiolucent 0.94 ± 0.35 1.59 ± 0.64 0.78 ± 0.31 2.10 ± 0.97 1.25 ± 0.46 1.48 ± 0.64 1.68 ± 0.45 1.31 ± 0.49 X-marker Donut marker 1.38 ± 0.68 1.69 ± 0.73 1.80 ± 0.85 4.21 ± 1.94 1.43 ± 0.56 2.20 ± 0.85 2.74 ± 1.19 2.77 ± 1.18 Anatomical 1.62 ± 0.65 1.77 ± 0.67 1.58 ± 0.52 2.69 ± 1.08 1.53 ± 0.49 1.94 ± 0.67 2.57 ± 1.20 2.02 ± 0.62 landmarks The worst-case TRE for group A and skin-affixed markers is 1.77 ± 0.71 mm (1.29 ± 0.41 mm radiolucent markers and 2.26 ± 0.92 mm donut markers) and 2.10 ± 0.83 mm for anatomical landmarks (Table 5). Group B has a worst-case TRE of 2.41 ± 0.81 mm for the anatomical landmarks and 2.42 ± 1.09 mm (1.75 ± 0.73 radiolucent markers and 3.10 ± 1.35 mm donut markers). For the worst-case scenario, no significant changes occurred to the statistical results and only the p value for A4–B4 increased to 0.43. Discussion This study evaluated four different registration methods using two different pointers. From a first point of view, it was Fig. 7 Pointwise TRE (in mm). On each target, the TRE was calculated apparent that the microscope led to larger TREs than the over all registration methods. The black bar is the mean TRE of target 1–10 when the probe is used (A1–A4), the white bar is the mean TRE probe. However, with additional surface registration plus for target 1–10 when the microscope is used (B1–B4). Targets 1–10 are using the VBH mouthpiece, the probe registration led to shown and described in Fig. 3 larger TREs than the microscope. Statistically no difference between microscope and probe registration could be found in this case. In general, more errors are influencing the accuracy for the microscope; in y-direction TRE is 1.08 ± 0.60 mm, of the microscope compared to the probe. Both pointers are in x-direction 1.31 ± 0.80 mm and in z direction 1.27 ± tracked, but the error chain of the microscope also includes 0.72 mm. camera calibration transformation errors. From this aspect, The TRE of anatomical targets and skin-affixed markers it can also be argued that the accuracy of the microscope is can be seen in Table 3. Skin-affixed markers have a TRE of similar to that of the probe. 1.50 ± 0.67 mm for group A and 2.16 ± 1.02 mm for group In general, surface registration is more time intense than B. The TRE for the anatomical landmarks is 1.82 ± 0.77 mm paired-point registration. But also locating the fiducials and (group A) and 2.10 ± 0.78 mm (group B). targets with the microscope takes much more time due to dif- Dividing skin-affixed markers in radiolucent and donut ficulties in focusing the points. Those two facts were already markers, the TRE of group A is 1.16 ± 0.40 and 1.84 ± speaking against using the microscope and an additional sur- 0.85 mm, respectively, and of group B 1.62 ± 0.71 mm and face registration before any calculations were done. However, 2.71 ± 1.27 mm, respectively. it is evident that some interventions, e.g. at the petrous bone, Pointwise evaluation is shown in Fig. 7. Target number 10 cannot be done without a navigated microscope and thus the (right infraorbital foramen) is the only point, where using the time factor is negligible. probe leads to a larger TRE than using the microscope. The worst results can be observed for B2 and A4, i.e. The results of the worst-case scenario can be seen in paired-point combined with surface registration is less accu- Tables 4, 5. The TRE for group A is 1.98 ± 0.80 mm which rate/precise than paired-point registration only. This result means a deterioration of the mean of 0.29 mm in comparison was also found in [14]. with the overall result. For group B, the worst-case TRE is However, there was no (statistically significant) difference 2.49 ± 0.99 mm; this is a deterioration of the mean of 0.30 between A1, A2 and A3. All those methods led to a mean mm compared to the overall TRE. TRE lower than 1.5 mm. Though it has to be said that A1 was 123 1546 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 Table 4 Worst-case scenario. TREs with the p values of the Wilcoxon signed-rank test TRE (mm) A1 B1 A2 B2 A3 B3 A4 B4 Mean ± std (worst case) 1.69 ± 0.63 1.93 ± 0.71 1.83 ± 0.68 3.37 ± 1.46 1.68 ± 0.52 2.22 ± 0.74 2.74 ± 1.19 2.43 ± 0.85 p value 0.1934 0.0020 0.0195 0.4315 It can be seen that A2–B2 and A3–B3 are statistically not equal (H : A = B , i = 1,..., 4). Mean values and standard deviations are given in mm 0 i i Table 5 Worst-case scenario: Mean TRE for the different landmarks used as targets for all registration methods TRE (mm) Mean ±stdA1 B1A2 B2A3 B3A4 B4 Radiolucent X-marker 1.05 ± 0.36 1.67 ± 0.67 0.89 ± 0.33 2.25 ± 1.01 1.45 ± 0.46 1.61 ± 0.65 1.77 ± 0.47 1.45 ± 0.49 Donut marker 1.75 ± 0.69 1.94 ± 0.78 2.36 ± 1.00 4.66 ± 2.06 1.78 ± 0.56 2.55 ± 0.91 3.13 ± 1.26 3.26 ± 1.26 Anatomical landmarks 1.90 ± 0.68 2.03 ± 0.68 1.89 ± 0.53 3.05 ± 1.15 1.71 ± 0.52 2.26 ± 0.66 2.90 ± 1.33 2.32 ± 0.64 the method that needed the fewest preparations and was ‘easy The effect of the location of the fiducials that are used for to use’. Thus, paired-point registration only is an adequate registration is an interesting point which can influence the method for our experiments. accuracy of navigation. The fiducials for paired-point regis- Regarding the registration with the microscope, B1, B3 tration are the same for A1/B1 and A2/B2, but also for A3/B3 and B4 were statistically equal. Using the microscope was and A4/B4. As shown in the results only A4 and B2 are sig- more time-consuming due to the waiting time until the micro- nificantly different to the other groups. But both methods scope is at rest and the autofocus is adjusted to confirm a are using surface registration in addition that should improve fiducial/target. However, when only paired-point registration results, especially as a rigid surface was used [21]. The stan- was applied, the time factor can be neglected and the TRE dard deviations for both A4 and B2 are about the double of the was similar to the experiments done with the probe. standard deviations of A1, A2, A3 and B1, B3, B4, respec- Pointwise evaluation clearly showed that donut markers tively. Registration, measurement or user errors could be the were the worst targets to use. This may be caused by the form cause for this discrepancy. Besides that, the difference seems of this marker, which is difficult to focus with the microscope. to be caused due to difficulties in locating the targets. Thus, It is also possible that those markers were difficult to locate it seems that external fiducials on the VBH mouthpiece do in the CT images, or the centre might not have been clearly influence the accuracy of navigation in a negative way when visible. For our experiments, it would have been better not to a microscope is used. use donut markers as a target. Anatomical landmarks showed Analysing the worst-case scenario, it can be observed that better TREs, the smallest values could be obtained at X-spots. the deterioration of the mean TRE is in all cases not more The targets with the largest mean TRE when using the probe than 0.35 mm, the standard deviation staying quite constant were the anatomical marker target nr. 9, and the Donut marker for all cases. This means that measurements were good and target nr. 5 when the microscope was used. constant, without large outliers. Considering the different It is clear that on a plastic skull anatomical landmarks types of targets, the values of the X-spots increase about and skin affixed markers can be located more accurately 0.1 mm, but Donut markers have mean values more than 0.5 in contrast to a real patient, where the skin is moving, or mm worse. Anatomical landmarks have a mean TRE that is bony structures, like foramina cannot be touched precisely. about 0.4 mm worse than in the overall case. This speaks Besides that, it is hard to realize not using anatomical land- again against using Donut markers but also against anatom- marks, because they are easy to use, non-invasive and clinical ical landmarks, because locating them is difficult and leads standard. to a large error. In contrary to the overall scenario, target 5 Anisotropy of the TRE in x-, y- and z-directions could be has the largest mean TRE for both groups A and B in the detected. In our experiments, the mean TRE is about 0.3 mm worst-case scenario. This target is a Donut marker on the left smaller in y-direction than in x- and z-directions. Typically, parietal bone. in optical systems the viewing direction has the largest noise Overall the largest TREs can be observed on the targets 5 variance, e.g. [20]. The anisotropy found in our set-up can be and 9 for both groups (Donut marker on the left parietal bone caused by not orienting the probe/microscope correctly to the and the saddle point on the right frontozygomatic suture), the camera, the underlying FLE, or the variance of the camera. smallest TREs on targets 1 and 2 (X-spots). Due to the exper- Besides that, the slice thickness of the CT images is larger imental setting, locating target 5 might be difficult (Fig. 8). than the pixel resolution of the images. Target 9 is a very unspecific point and is difficult to locate 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1547 Conclusion The target registration error of different registration point- ers and methods was evaluated, and differences between the alternative approaches could be observed. Overall it can be observed that a probe used with paired-point registration gives the best results for navigated surgery. Additional sur- face registration and/or fiducials do not improve the TRE significantly in our experiments. Using a microscope as a pointer leads to a larger error compared to the probe for all experiments except when paired-point registration is combined with external fiducials and surface registration. However, in clinical practice pointers are mostly used for registering the patient to pre- and intraoperative radiological data. In conclusion, both registration instruments had com- parable application accuracies for the plastic skull in an experimental set-up. Using the microscope is a rarely used registration device, but this study demonstrates that registra- tion outcome is not affected by the registration tool chosen. Acknowledgements Open access funding provided by University of Innsbruck and Medical University of Innsbruck. Funding This project was funded by the Jubilee Funds of the Austrian Fig. 8 Plastic skull with the VBH mouthpiece fixed with the VBH National Bank, Projects Nos. 13003, 14751, and 16903 and by the Aus- holder to the OR table. The DRF can be seen on the right side of the trian Research Promotion Agency (FFG) under project number 846056. plastic skull [23] Compliance with ethical standards both in the CT image and on the plastic skull, in contrary to targets 1 and 2, which are X-spots, that are easy to locate Conflict of interest The authors declare that they have no conflict of interest. on the front of the skull. X-spots can be located more pre- cisely than all other markers, they have the smallest standard Ethical approval This article does not contain any studies with human deviations. In contrary to that, Donut markers have the largest participants or animals performed by any of the authors. standard deviations; only in experiments A4 and B4, anatom- Informed consent This articles does not contain patient data. ical landmarks can be located less precisely. Unfortunately, the study is limited to a paired-point reg- Open Access This article is distributed under the terms of the Creative istration using 10 fiducials throughout the experiments. As Commons Attribution 4.0 International License (http://creativecomm already confirmed by other studies (e.g. see [8,10,22]), a ons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit smaller number of fiducials lead to a larger TRE than using to the original author(s) and the source, provide a link to the Creative more fiducials for registration. However, using, for example, Commons license, and indicate if changes were made. 5 fiducials is not necessarily leading to a larger TRE than using 10 fiducials. It is always depending on the situation and on the fiducials used. Due to not using a different num- References ber of fiducials for the same experiment, we do not know if this would lead to an improvement for one of the registration 1. Horn BKP (1987) Closed-form solution of absolute orientation methods. However, the 10 fiducials used were placed ideally using unit quaternions. J Opt Soc Am A 4(4):629–642 on the plastic skull to ensure a good configuration for the 2. Besl PJ, McKay HD (1992) A method for registration of 3-D registration process. shapes. IEEE Trans Pattern Anal Mach Intell 14(2):239–256 3. Citardi M (2017) Image-guided surgery: Fundamentals and clinical applications in otolaryngology Robert Labadie, J. Michael Fitz- patrick Plural Publishing, San Diego CA, 2016, USD 149.95, vol 39, 215 pp 4. Bale RJ, Vogele M, Martin A, Auer T, Hensler E, Eichberger P, Freysinger W, Sweeney R, Gunkel AR, Lukas PH (1997) VBH 123 1548 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 head holder to improve frameless stereotactic brachytherapy of 15. Schicho K, Figl M, Seemann R, Donat M, Pretterklieber ML, Birk- cranial tumors. Comput Aided Surg 2(5):286–291 fellner W, Reichwein A, Wanschitz F, Kainberger F, Bergmann H, 5. Sweeney RA, Bale R, Auberger T, Vogele M, Foerster S, Nevinny- Wagner A, Ewers R (2007) Comparison of laser surface scanning Stickel M, Lukas P (2001) A simple and non-invasive vacuum and fiducial marker-based registration in frameless stereotaxy. J mouthpiece-based head fixation system for high precision radio- Neurosurg 106(4):704–709 therapy. Strahlentherapie und Onkol 177(1):43–47 16. Sun Y, Luebbers H-T, Agbaje JO, Schepers S, Vrielinck L, 6. Martin A, Bale RJ, Vogele M, Gunkel AR, Thumfart WF, Lambrichts I, Politis C (2013) Validation of anatomical landmarks- Freysinger W (1998) Vogele–Bale–Hohner mouthpiece: reg- based registration for image-guided surgery: an in-vitro study. J istration device for frameless stereotactic surgery. Radiology Craniomaxillofac Surg 41(6):522–526 208(1):261–265 17. Hardy SM, Melroy C, White DR, Dubin M, Senior B (2006) A 7. Bardosi Z, Freysinger W (2016) Estimating FLE_image distribu- comparison of computer-aided surgery registration methods for tions of manual fiducial localization in CT images. Int J Comput endoscopic sinus surgery. Am J Rhinol 20(1):48–52 Assist Radiol Surg 11(6):1043–1049 18. Mascott CR, Sol J-C, Bousquet P, Lagarrigue J, Lazorthes Y, 8. Perwög M, Bardosi Z, Freysinger W (2018) Experimental vali- Lauwers-Cances V (2006) Quantification of true in vivo (applica- dation of predicted application accuracies for computer-assisted tion) accuracy in cranial image-guided surgery: influence of mode (CAS) intraoperative navigation with paired-point registration. Int of patient registration. Neurosurgery 59(1 Suppl 1):46–56 J Comput Assist Radiol Surg 13(3):425–441 19. Metzger MC, Rafii A, Holhweg-Majert B, Pham AM, Strong B 9. Fitzpatrick JM (2010) The role of registration in accurate surgical (2007) Comparison of 4 registration strategies for computer-aided guidance. Proc Inst Mech Eng H 224(5):607–622 maxillofacial surgery. Otolaryngol Neck Surg 137(1):93–99 10. Güler Ö, Perwög M, Kral F, Schwarm F, Bárdosi ZR, Göbel 20. Ma B, Moghari MH, Ellis RE, Abolmaesumi P (2010) Estima- G, Freysinger W (2013) Quantitative error analysis for computer tion of optimal fiducial target registration error in the presence of assisted navigation: a feasibility study. Med Phys 40(2):21910 heteroscedastic noise. IEEE Trans Med Imaging 29(3):708–723 11. Sonka M, Fitzpatrick JM (2000) Handbook of medical imaging, vol 21. Maurer CR, Maciunas RJ, Fitzpatrick JM (1998) Registration of 2. Medical image processing and analysis. SPIE The international head CT images to physical space using a weighted combination society for optical engineering, Bellingham of points and surfaces [image-guided surgery]. IEEE Trans Med 12. Fitzpatrick JM (2009) Fiducial registration error and target regis- Imaging 17(5):753–761 tration error are uncorrelated. In: Proceedings of SPIE, vol 7261 22. Kim S, Kazanzides P (2017) Fiducial-based registration with 13. Luebbers H-T, Messmer P, Obwegeser JA, Zwahlen RA, Kikinis a touchable region model. Int J Comput Assist Radiol Surg R, Graetz KW, Matthews F (2008) Comparison of different reg- 12(2):277–289 istration methods for surgical navigation in cranio-maxillofacial 23. Jeleff O (2005) Mikroskop- und pointerbasierte Navigation an der surgery. J Cranio-Maxillofac Surg 36(2):109–116 Laterobasis, MD-Thesis, Medizinische Universität Innsbruck 14. Woerdeman PA, Willems PWA, Noordmans HJ, Tulleken CAF, van der Sprenkel JWB (2007) Application accuracy in frameless image-guided neurosurgery: a comparison study of three patient- to-image registration methods. J Neurosurg 106(6):1012–1016 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Computer Assisted Radiology and Surgery Springer Journals

Probe versus microscope: a comparison of different methods for image-to-patient registration

Free
10 pages

Loading next page...
 
/lp/springer_journal/probe-versus-microscope-a-comparison-of-different-methods-for-image-to-m2SUzMtYId
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Imaging / Radiology; Surgery; Health Informatics; Computer Imaging, Vision, Pattern Recognition and Graphics; Computer Science, general
ISSN
1861-6410
eISSN
1861-6429
D.O.I.
10.1007/s11548-018-1800-0
Publisher site
See Article on Publisher Site

Abstract

Purpose Computer-aided navigation is widely used in ENT surgery. The position of a surgical instrument is shown in the CT/MR images of the patient and can thus be a good support for the surgeon. The accuracy is highly dependent on the registration done prior to surgery. A microscope and a probe can both be used for registration and navigation, depending on the surgical intervention. A navigation system typically only reports the fiducial registration error after paired-point registration. However, the target registration error (TRE)—a measurement for the accuracy in the surgical area—is much more relevant. The aim of this work was to compare the performance of a microscope relative to a conventional probe-based approach with different registration methods. Methods In this study, optical tracking was used to register a plastic skull to its preoperative CT images with paired-point registration. Anatomical landmarks and skin-affixed markers were used as fiducials and targets. With both microscope and probe, four different registration methods were evaluated based on their TREs at 10 targets. For half of the experiments, a surface registration and/or external fiducials were used additionally to paired-point registration to study their influence to accuracy. Results Overall, probe registration leads to a smaller TRE (1.69 ± 0.74 mm) than registration with a microscope (2.19 ± 0.94 mm). Additional surface registration does not result in better accuracy of navigation for microscope and probe. The lowest mean TRE for both pointers can be achieved with paired-point registration only and radiolucent markers. Conclusion Our experiments showed that a probe used for registration and navigation achieves lower TREs compared using a microscope. Neither additional surface registration nor additional fiducials on an external reference element are necessary for improved accuracy of navigated ENT surgery on a plastic skull. Keywords Registration · Microscope · Navigation · Target registration error Introduction skull base to support the surgeon with detailed images of the surgical area. In this case, the pointer can be the microscope Clinical navigation systems are commonly used for ENT or a probe. interventions. At critical regions, they are supporting the Prior to using a navigation system, the CT/MR images surgeon by realizing the exact positions of a pointer inside have to be registered to the patient. In general, the two main the patients’ body on the CT/MR images. Depending on the registration methods are paired-point and surface registration surgical intervention, different pointers can be used for nav- [1,2]. For the first method, points on the patient (‘fiducials’) igation; for example, a microscope is required at the lateral have to be located, together with the corresponding points in the images. Thus, the rigid transformation between image and patient coordinates can be calculated. Besides, anatom- B Martina Perwög martinaperwoeg@hotmail.com ical structures also skin-affixed or bone-implanted markers can be used as fiducials. Skin-affixed markers are known to ENT/4D-Visualization, Medical University Innsbruck, be prone to changes of the skin according to different hydra- Anichstr. 35, Innsbruck, Austria tion statuses or effects of muscle relaxing agents and thus are ENT, Klinikum der Universität München, Munich, Germany to be used with caution [3]. Those aspects can be avoided if ENT, Kardinal Schwarzenberg Klinikum, Schwarzach im bone-implanted fiducials are used, but because of the inva- Pongau, Austria 123 1540 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 Table 1 Overview of the different experiments with the number of fiducials and targets used for registrations and measurements Pointer # Fiducials # Targets Probe Microscope Anat. Donut Radiolucent Surface reg. VBH Anat. Donut Radiolucent A1 B1 5 2 3 – – 5 3 2 A2 B2 5 2 3 25 – A3 B3 3 1 2 – 4 A4 B4 3 1 2 25 4 siveness they are not the first choice, due to side effects for the patient. The most commonly used fiducials are anatomi- cal landmarks, as they are easy to use and non-invasive. Surface registration helps automation and registers a larger number of points to the image. These points are collected by moving a pointer over a large surface on the patient, like the forehead. With the iterative-closest-point (ICP) algorithm, the surface of the patient is matched to the corresponding 3D-image-model [2]. For interventions, e.g. the lateral skull base or the petrous bone, the VBH (Vogele–Bale–Hohner) head holder can be used additionally to fix the patients head on the OR Table [4]. External fiducials can be affixed to the VBH mouthpiece to get a larger area of fiducials that surrounds the surgical Fig. 1 Experimental set-up for experiments A1,…, A4. The plastic region and thus improve accuracy [5,6]. More details are skull is fixed with the VBH head holder on the OR table. The camera given in the ‘Materials and methods’ section. is optimally placed and the navigation system placed opposite to the The ideal case for the surgeon would of course be per- surgeon [23] fect (error free) navigation, but the registration is a critical step and has to be done as accurately as possible. However, Two different pointers were used for both registration and it always introduces errors as calibration, tracking and mea- TRE measurement: a microscope and a pointer. Both pointers surement errors can occur, but user localization errors are are optically tracked by a camera relative to a DRF (digital also common in surgical navigation [7–10]. reference frame), so the coordinates of the points in patient After registration, the typical navigation systems report space can be determined and transformed to image space. a single error which is called the fiducial registration error Four different registration methods have been experimen- (FRE) [11]. The FRE is the RMS distance of CT image fidu- tally evaluated with each pointer (probe or microscope), two cials and patient fiducials that are transformed to the CT of them with external fiducials on the VBH mouthpiece. image. However, the target registration error (TRE) is much Surface registration was additionally used for half of the more important for the surgeon in practice. It reports the experiments (see Table 1 for a detailed description). The TRE accuracy at a target location close to the surgically relevant was measured on 10 different targets, and all methods were positions and is uncorrelated with the FRE [12]. compared to each other. Thus, it was possible to evaluate the Unfortunately, not many studies regarding the applica- performance and differences of the registration methods and tion accuracy (the TRE) of different registration methods pointers used for our experiments. have been done comparing microscope and probe registra- tion methods [13–19]. Therefore, the aim of this study was to find the optimal Materials and methods registration method of a plastic skull to its preoperative CT images in terms of application accuracy at the lateral skull The experimental set-up can be seen in Figs. 1, 2. base, concerning the pointer and the registration method. Two different questions were evaluated: What kind of pointer Phantom should be used for registration and measurements? Which registration method is the optimal method to register a plas- A plastic skull (Somso, Coburg, Germany) was used for all tic skull to its CT images? experiments. Radiolucent markers (X-spot, izi-Kooperation, 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1541 Fig. 3 Plastic skull with the VBH mouthpiece. Targets are shown in pink; fiducials are shown in black. The number in green shows the Fig. 2 Experimental set-up when the microscope is used for the regis- fiducial used for registration only with VBH mouthpiece. (fiducials: tration. The DRF is fixed on the left side of the microscope (the blue 3—X-spot on the frontal bone, 4—X-spot on the left nasal bone, 6— cylinder hidden in this figure on the left side) [23] anatomical landmark on the right supraorbital foramen, 7—right inner cantus, 8—anterior nasal spine, 9—saddle point on the left frontozy- gomatic suture, 10—infraorbital foramen; targets: 1, 2—radiolucent markers on the right frontal bone and the right maxillary bone, 3— Beekley Corp., Conn, USA) and MR marker (Donuts, izi- donut marker on the left frontal bone, 6—left supraorbital foramen, Kooperation, Beekley Corp., Conn., USA) were attached on 7—left inner canthus, 8—left nasal bone, 9—saddle point right fron- tozygomatic suture, 10—right infraorbital foramen) [23] the skull. For all experiments the plastic skull was fixed on an OR table with a VBH head holder [4]. With the VBH mouth- piece, the skull can be attached at the holder [6]. Dental foam Navigation system vacuum fixation on the maxilla leads to a stable construction in all experiments. This is a fast and secure non-invasive The SNN navigation system (Fa. SNS, Aalen, Germany) was method to fix a patient during interventions. On the refer- used for navigation with optical tracking of (flashpoint-3000- ence element of the VBH mouthpiece, radiolucent markers system, IGT, Boulder, USA). It comprises classical pointer- were attached (see Figs. 3, 4). Those external markers were based navigation and a navigated microscope with heads-up only used for half of the experiments (see Table 1). display capabilities for navigation. Pointers CT images Two different pointers were used for registration and naviga- CT images of the plastic skull were acquired with a Siemens- tion on a plastic skull. The probe (135-mm flashpoint) can be Sensation-16 (Siemens, Erlangen, Germany), leading to 2 optically tracked by the camera with the two LEDs attached axial datasets. The first dataset was captured with the refer- to it and is connected with a cable to the navigation system. ence element fixed on the VBH mouthpiece in place and the The microscope (OPMI 2000, Zeiss, Oberkochen, Ger- second one without the reference element. Image parameters many) can be optically tracked with a DRF attached to it (see were 140 kV, 220 mAs with 1 mm slice thickness. Fig. 2). The microscope is factory-calibrated, and thus, the 123 1542 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 and the microscope was not moving anymore, this point was used for registration. In detail (cf. Figs. 3, 4), the fiducials used for the experi- ments without VBH mouthpiece are the 2 donut markers on the right parietal bone and the right frontal bone, the 3 radi- olucent markers on the frontal bone, the left nasal bone and the left temporal bone. The five anatomical landmarks are the right supraorbital foramen, the right medial cantus, the anterior nasal spine, the ‘saddle point’ on the left frontozy- gomatic suture, and the left infraorbital foramen. For the experiments with the VBH mouthpiece, the fidu- cials in detail are 4 radiolucent markers on the referencing element of the VBH mouthpiece, 1 donut marker at the right parietal bone, 2 radiolucent markers median on the frontal Fig. 4 Plastic skull from the right with fiducials (black) and fiducials on bone and the left temporal bone and the 3 anatomical land- the VBH mouthpiece (green). (1—Donut marker on the right parietal marks on the anterior nasal spine, the left infraorbital foramen bone, 2—Donut marker on the right frontal bone, 3 and 4—X-spots on the VBH—mouthpiece used only for registration with the VBH and the right supraorbital foramen. mouthpiece) [23] For the surface registration, 25 points were acquired by the probe touching the mid-face, the frontal bone and the orbitae of the skull surface. The same anatomical locations were parameters of the camera are known and the position of the focused with the microscope with maximum magnification. point that is in focus can be determined. Again 25 points were selected, when the points were in focus and the microscope was not moving. Registration TRE measurements To register the plastic skull to its CT images, four different After all different registration methods, the TRE was mea- registration methods were performed with the probe (group sured on 10 targets, which were per definition different to the A) or the microscope (group B). As a skull has only a few fiducials (c.f. Fig. 3). The targets used were always the same prominent anatomical landmarks that can be defined well for A1–A4 and B1–B4. both on the patient and in the CT images, different kinds Five anatomical landmarks, 2 radiolucent markers and 3 of markers were used as fiducials (but also targets, as can donut markers were used. In detail, the targets (in ascending be seen in the next section), to have a sufficient amount of order target number 1–10) are the 2 radiolucent markers on fiducials for each registration method. the right frontal bone and the right cheek bone, the donut The different methods and fiducials used for registration markers on the left frontal bone, near the left coronal suture were defined as follows (see also Table 1): and the left parietal bone. The 5 anatomical targets are the A1 and B1: paired-point registration with 5 anatomical left supraorbital foramen, the left medial cantus, the bony landmarks, 2 donut markers and 3 radiolucent markers structure on the left nasal bone, the ‘saddle point’ on the right A2 and B2: paired-point registration with 5 anatomical frontozygomatic suture and the right infraorbital foramen. landmarks, 2 donut markers, 3 radiolucent markers and addi- Each experiment (registration and TRE measurement) was tionally a surface registration with 25 points repeated 10 times. To avoid a learning effect during the A3 and B3: paired-point registration with 3 anatomical experiments, initial experiments were done to get used to landmarks, 1 donut marker and 2 radiolucent markers on the the equipment without recording the results. skull and 4 radiolucent markers on the VBH mouthpiece Each target was located as exactly as possible with the A4 and B4: paired-point registration with 3 anatomical probe. A screenshot of the navigation systems display was landmarks, 1 donut marker and 2 radiolucent markers on the taken, when the probe was placed at the target. The same skull, 4 radiolucent markers on the VBH mouthpiece and procedure was repeated with the microscope, and each tar- additionally a surface registration with 25 points get was located with the maximum magnification and with The probe and the microscope were both used as point- autofocus. ers to locate a fiducial on the plastic skull. The fiducials After finishing the experiments, each screenshot was pro- were physically located with the tip of the probe. With the cessed in Adobe Photoshop 5.0. Each image plane was microscope, the fiducials were located with maximum mag- maximally magnified. The difference of real target and pro- nification. Only when the fiducial of interest was in autofocus jected target (Δx, Δy, Δz) was measured in each image 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1543 Fig. 5 Screenshot of the navigation system’s monitor when pointing on the X-spot. The difference between the target point and the pointer tip is measured in axial, coronal and sagittal views (c.f. Fig. 6, right). The crosshair is used to get the scaling for the measurement (c.f. Fig. 6,left) [23] 2 2 2 plane, axial (x–y), sagittal (x–z) and coronal (y–z). So each TRE(t ) = x + y + z , ij ij ij ij x-, y- and z direction was measured twice (10 × 6 data matrix for each group). Those pixel measurements were converted which is the TRE for target number i, i = 1,..., 10, experi- in mm. The scale is known by measuring the length of the ment j, j = A1,..., A4, B1,..., B4. crosshair, which is 1 cm. An example of the TRE measure- All calculations were done with all measurements of each ment is shown in Figs. 5, 6. group in Matlab. It could be possible that the target was not visible in one Besides that, a worst-case scenario was analysed, where plane, and thus, the error could only be measured in the other only the largest values in each x-, y-, and z-directions of each two planes. The missing values were assigned as NaN (not-a- repetition were used for calculations (10 × 3 data matrix for number) and replaced with the corresponding value in x-, y- each group). and z-directions; for example, if the axial x-direction could not be measured, the measurement of the sagittal x- direction Evaluation was used for both x-values. Overall 4800 measurements were taken, 600 for each The aim of this investigation was to find the difference regard- experiment. For each target, the mean of the 10 repetitions in ing the TRE between registrations via microscope and probe each direction was calculated. Thus, the values (x , y , ij ij on a plastic skull. Besides that, it was evaluated if surface z ) remained for each target. With these values, the TRE ij registration (additionally to paired-point registration) and was calculated by additional fiducials fixed on the reference element of the 123 1544 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 Fig. 6 Left side: measuring the length of the crosshair, which is defined the navigation system. The probe (blue) is pointing at the target (white as 1 cm. The measured value is 1.62 in this image and is used as the elliptical point). The difference between the tip of the probe and the scaling factor for the measurements. Right side: example of the TRE centre of the target is 0.05 [23] measurement on a X-spot. Maximized part of the image of the display of Table 2 TREs with the p values of the Wilcoxon signed-rank test TRE (mm) A1 B1 A2 B2 A3 B3 A4 B4 Mean ± std 1.41 ± 0.61 1.70 ± 0.68 1.48 ± 0.61 3.02 ± 1.37 1.44 ± 0.50 1.92 ± 0.72 2.44 ± 1.09 2.10 ± 0.81 p value 0. 0645 0.0020 0. 0273 0.1602 It can be seen that A2–B2 and A3–B3 are statistically not equal. (H :A = B , i = 1,..., 4). Mean values and standard deviations are given in mm 0 i i VBH mouthpiece improve the TRE. Group A (probe) and ference of microscope- and probe-based registration can also group B (microscope) were compared to each other. The over- be observed (Wilcoxon signed rank, p < 0.0001). all TRE over all four types of registration was calculated for a The overall TRE for group A is 1.69 ± 0.74 mm, and for first overview, but also the TRE of each registration method group B, it is 2.19 ± 0.94 mm. was calculated separately. All calculations were done with The mean TRE of A1 is 1.41 mm compared to B1 with Matlab R2012a. Statistical tests were done with α = 0.05. a mean TRE of 1.70 mm. The mean TRE of A2 is 1.48 mm An ANOVA was used to find differences in group A or group compared to B2 with a mean TRE of 3.02. A3 and B3 have B. If a statistically significant difference was found, multiple a mean TRE of 1.44 mm and 1.92 mm, respectively. A4 and comparisons were done with Bonferroni correction to anal- B4 have a mean TRE of 2.44 mm and 2.10 mm, respectively yse which experiments were significantly different to each (Table 2). other. For comparing Ai and Bi, i = 1,..., 4, a Wilcoxon These results are leading to statistical results as follows: signed-rank test was used. A1 is equal to B1 ( p = 0.0645) and A4 is equal to B4 ( p = 0.1602). A2 and B2 are significantly different ( p = 0.0020), as well as A3 and B3 ( p = 0.0273). The four probe groups (A1–A4) are significantly different Results to each other ( p = 0.0004). A4 has the largest error. B1–B4 are significantly different to each other ( p = All results are mean values with their standard deviation and 0.0012), and B2 has the largest TRE. can be seen in detail in Tables 2, 3. Group A defines the Regarding the error in x-, y- and z-directions, it could be registration with the probe and group B the registration with observed that overall the TRE is anisotropic. For the probe the microscope. measurements, the error in y-direction is the smallest with All errors in the groups, except A1, A2 and B2, are nor- 0.79 ± 0.49 mm and in x- and z-directions the error is 1.00 ± mally distributed ( p < 0.05). All 8 groups compared to each 0.56 mm and 1.03 ± 0.62 mm, respectively. This is also valid other are not equal (ANOVA, p < 0.0001); an overall dif- 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1545 Table 3 Mean TRE at different landmarks used as targets for the 8 groups TRE (mm) A1 B1 A2 B2 A3 B3 A4 B4 Radiolucent 0.94 ± 0.35 1.59 ± 0.64 0.78 ± 0.31 2.10 ± 0.97 1.25 ± 0.46 1.48 ± 0.64 1.68 ± 0.45 1.31 ± 0.49 X-marker Donut marker 1.38 ± 0.68 1.69 ± 0.73 1.80 ± 0.85 4.21 ± 1.94 1.43 ± 0.56 2.20 ± 0.85 2.74 ± 1.19 2.77 ± 1.18 Anatomical 1.62 ± 0.65 1.77 ± 0.67 1.58 ± 0.52 2.69 ± 1.08 1.53 ± 0.49 1.94 ± 0.67 2.57 ± 1.20 2.02 ± 0.62 landmarks The worst-case TRE for group A and skin-affixed markers is 1.77 ± 0.71 mm (1.29 ± 0.41 mm radiolucent markers and 2.26 ± 0.92 mm donut markers) and 2.10 ± 0.83 mm for anatomical landmarks (Table 5). Group B has a worst-case TRE of 2.41 ± 0.81 mm for the anatomical landmarks and 2.42 ± 1.09 mm (1.75 ± 0.73 radiolucent markers and 3.10 ± 1.35 mm donut markers). For the worst-case scenario, no significant changes occurred to the statistical results and only the p value for A4–B4 increased to 0.43. Discussion This study evaluated four different registration methods using two different pointers. From a first point of view, it was Fig. 7 Pointwise TRE (in mm). On each target, the TRE was calculated apparent that the microscope led to larger TREs than the over all registration methods. The black bar is the mean TRE of target 1–10 when the probe is used (A1–A4), the white bar is the mean TRE probe. However, with additional surface registration plus for target 1–10 when the microscope is used (B1–B4). Targets 1–10 are using the VBH mouthpiece, the probe registration led to shown and described in Fig. 3 larger TREs than the microscope. Statistically no difference between microscope and probe registration could be found in this case. In general, more errors are influencing the accuracy for the microscope; in y-direction TRE is 1.08 ± 0.60 mm, of the microscope compared to the probe. Both pointers are in x-direction 1.31 ± 0.80 mm and in z direction 1.27 ± tracked, but the error chain of the microscope also includes 0.72 mm. camera calibration transformation errors. From this aspect, The TRE of anatomical targets and skin-affixed markers it can also be argued that the accuracy of the microscope is can be seen in Table 3. Skin-affixed markers have a TRE of similar to that of the probe. 1.50 ± 0.67 mm for group A and 2.16 ± 1.02 mm for group In general, surface registration is more time intense than B. The TRE for the anatomical landmarks is 1.82 ± 0.77 mm paired-point registration. But also locating the fiducials and (group A) and 2.10 ± 0.78 mm (group B). targets with the microscope takes much more time due to dif- Dividing skin-affixed markers in radiolucent and donut ficulties in focusing the points. Those two facts were already markers, the TRE of group A is 1.16 ± 0.40 and 1.84 ± speaking against using the microscope and an additional sur- 0.85 mm, respectively, and of group B 1.62 ± 0.71 mm and face registration before any calculations were done. However, 2.71 ± 1.27 mm, respectively. it is evident that some interventions, e.g. at the petrous bone, Pointwise evaluation is shown in Fig. 7. Target number 10 cannot be done without a navigated microscope and thus the (right infraorbital foramen) is the only point, where using the time factor is negligible. probe leads to a larger TRE than using the microscope. The worst results can be observed for B2 and A4, i.e. The results of the worst-case scenario can be seen in paired-point combined with surface registration is less accu- Tables 4, 5. The TRE for group A is 1.98 ± 0.80 mm which rate/precise than paired-point registration only. This result means a deterioration of the mean of 0.29 mm in comparison was also found in [14]. with the overall result. For group B, the worst-case TRE is However, there was no (statistically significant) difference 2.49 ± 0.99 mm; this is a deterioration of the mean of 0.30 between A1, A2 and A3. All those methods led to a mean mm compared to the overall TRE. TRE lower than 1.5 mm. Though it has to be said that A1 was 123 1546 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 Table 4 Worst-case scenario. TREs with the p values of the Wilcoxon signed-rank test TRE (mm) A1 B1 A2 B2 A3 B3 A4 B4 Mean ± std (worst case) 1.69 ± 0.63 1.93 ± 0.71 1.83 ± 0.68 3.37 ± 1.46 1.68 ± 0.52 2.22 ± 0.74 2.74 ± 1.19 2.43 ± 0.85 p value 0.1934 0.0020 0.0195 0.4315 It can be seen that A2–B2 and A3–B3 are statistically not equal (H : A = B , i = 1,..., 4). Mean values and standard deviations are given in mm 0 i i Table 5 Worst-case scenario: Mean TRE for the different landmarks used as targets for all registration methods TRE (mm) Mean ±stdA1 B1A2 B2A3 B3A4 B4 Radiolucent X-marker 1.05 ± 0.36 1.67 ± 0.67 0.89 ± 0.33 2.25 ± 1.01 1.45 ± 0.46 1.61 ± 0.65 1.77 ± 0.47 1.45 ± 0.49 Donut marker 1.75 ± 0.69 1.94 ± 0.78 2.36 ± 1.00 4.66 ± 2.06 1.78 ± 0.56 2.55 ± 0.91 3.13 ± 1.26 3.26 ± 1.26 Anatomical landmarks 1.90 ± 0.68 2.03 ± 0.68 1.89 ± 0.53 3.05 ± 1.15 1.71 ± 0.52 2.26 ± 0.66 2.90 ± 1.33 2.32 ± 0.64 the method that needed the fewest preparations and was ‘easy The effect of the location of the fiducials that are used for to use’. Thus, paired-point registration only is an adequate registration is an interesting point which can influence the method for our experiments. accuracy of navigation. The fiducials for paired-point regis- Regarding the registration with the microscope, B1, B3 tration are the same for A1/B1 and A2/B2, but also for A3/B3 and B4 were statistically equal. Using the microscope was and A4/B4. As shown in the results only A4 and B2 are sig- more time-consuming due to the waiting time until the micro- nificantly different to the other groups. But both methods scope is at rest and the autofocus is adjusted to confirm a are using surface registration in addition that should improve fiducial/target. However, when only paired-point registration results, especially as a rigid surface was used [21]. The stan- was applied, the time factor can be neglected and the TRE dard deviations for both A4 and B2 are about the double of the was similar to the experiments done with the probe. standard deviations of A1, A2, A3 and B1, B3, B4, respec- Pointwise evaluation clearly showed that donut markers tively. Registration, measurement or user errors could be the were the worst targets to use. This may be caused by the form cause for this discrepancy. Besides that, the difference seems of this marker, which is difficult to focus with the microscope. to be caused due to difficulties in locating the targets. Thus, It is also possible that those markers were difficult to locate it seems that external fiducials on the VBH mouthpiece do in the CT images, or the centre might not have been clearly influence the accuracy of navigation in a negative way when visible. For our experiments, it would have been better not to a microscope is used. use donut markers as a target. Anatomical landmarks showed Analysing the worst-case scenario, it can be observed that better TREs, the smallest values could be obtained at X-spots. the deterioration of the mean TRE is in all cases not more The targets with the largest mean TRE when using the probe than 0.35 mm, the standard deviation staying quite constant were the anatomical marker target nr. 9, and the Donut marker for all cases. This means that measurements were good and target nr. 5 when the microscope was used. constant, without large outliers. Considering the different It is clear that on a plastic skull anatomical landmarks types of targets, the values of the X-spots increase about and skin affixed markers can be located more accurately 0.1 mm, but Donut markers have mean values more than 0.5 in contrast to a real patient, where the skin is moving, or mm worse. Anatomical landmarks have a mean TRE that is bony structures, like foramina cannot be touched precisely. about 0.4 mm worse than in the overall case. This speaks Besides that, it is hard to realize not using anatomical land- again against using Donut markers but also against anatom- marks, because they are easy to use, non-invasive and clinical ical landmarks, because locating them is difficult and leads standard. to a large error. In contrary to the overall scenario, target 5 Anisotropy of the TRE in x-, y- and z-directions could be has the largest mean TRE for both groups A and B in the detected. In our experiments, the mean TRE is about 0.3 mm worst-case scenario. This target is a Donut marker on the left smaller in y-direction than in x- and z-directions. Typically, parietal bone. in optical systems the viewing direction has the largest noise Overall the largest TREs can be observed on the targets 5 variance, e.g. [20]. The anisotropy found in our set-up can be and 9 for both groups (Donut marker on the left parietal bone caused by not orienting the probe/microscope correctly to the and the saddle point on the right frontozygomatic suture), the camera, the underlying FLE, or the variance of the camera. smallest TREs on targets 1 and 2 (X-spots). Due to the exper- Besides that, the slice thickness of the CT images is larger imental setting, locating target 5 might be difficult (Fig. 8). than the pixel resolution of the images. Target 9 is a very unspecific point and is difficult to locate 123 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 1547 Conclusion The target registration error of different registration point- ers and methods was evaluated, and differences between the alternative approaches could be observed. Overall it can be observed that a probe used with paired-point registration gives the best results for navigated surgery. Additional sur- face registration and/or fiducials do not improve the TRE significantly in our experiments. Using a microscope as a pointer leads to a larger error compared to the probe for all experiments except when paired-point registration is combined with external fiducials and surface registration. However, in clinical practice pointers are mostly used for registering the patient to pre- and intraoperative radiological data. In conclusion, both registration instruments had com- parable application accuracies for the plastic skull in an experimental set-up. Using the microscope is a rarely used registration device, but this study demonstrates that registra- tion outcome is not affected by the registration tool chosen. Acknowledgements Open access funding provided by University of Innsbruck and Medical University of Innsbruck. Funding This project was funded by the Jubilee Funds of the Austrian Fig. 8 Plastic skull with the VBH mouthpiece fixed with the VBH National Bank, Projects Nos. 13003, 14751, and 16903 and by the Aus- holder to the OR table. The DRF can be seen on the right side of the trian Research Promotion Agency (FFG) under project number 846056. plastic skull [23] Compliance with ethical standards both in the CT image and on the plastic skull, in contrary to targets 1 and 2, which are X-spots, that are easy to locate Conflict of interest The authors declare that they have no conflict of interest. on the front of the skull. X-spots can be located more pre- cisely than all other markers, they have the smallest standard Ethical approval This article does not contain any studies with human deviations. In contrary to that, Donut markers have the largest participants or animals performed by any of the authors. standard deviations; only in experiments A4 and B4, anatom- Informed consent This articles does not contain patient data. ical landmarks can be located less precisely. Unfortunately, the study is limited to a paired-point reg- Open Access This article is distributed under the terms of the Creative istration using 10 fiducials throughout the experiments. As Commons Attribution 4.0 International License (http://creativecomm already confirmed by other studies (e.g. see [8,10,22]), a ons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit smaller number of fiducials lead to a larger TRE than using to the original author(s) and the source, provide a link to the Creative more fiducials for registration. However, using, for example, Commons license, and indicate if changes were made. 5 fiducials is not necessarily leading to a larger TRE than using 10 fiducials. It is always depending on the situation and on the fiducials used. Due to not using a different num- References ber of fiducials for the same experiment, we do not know if this would lead to an improvement for one of the registration 1. Horn BKP (1987) Closed-form solution of absolute orientation methods. However, the 10 fiducials used were placed ideally using unit quaternions. J Opt Soc Am A 4(4):629–642 on the plastic skull to ensure a good configuration for the 2. Besl PJ, McKay HD (1992) A method for registration of 3-D registration process. shapes. IEEE Trans Pattern Anal Mach Intell 14(2):239–256 3. Citardi M (2017) Image-guided surgery: Fundamentals and clinical applications in otolaryngology Robert Labadie, J. Michael Fitz- patrick Plural Publishing, San Diego CA, 2016, USD 149.95, vol 39, 215 pp 4. Bale RJ, Vogele M, Martin A, Auer T, Hensler E, Eichberger P, Freysinger W, Sweeney R, Gunkel AR, Lukas PH (1997) VBH 123 1548 International Journal of Computer Assisted Radiology and Surgery (2018) 13:1539–1548 head holder to improve frameless stereotactic brachytherapy of 15. Schicho K, Figl M, Seemann R, Donat M, Pretterklieber ML, Birk- cranial tumors. Comput Aided Surg 2(5):286–291 fellner W, Reichwein A, Wanschitz F, Kainberger F, Bergmann H, 5. Sweeney RA, Bale R, Auberger T, Vogele M, Foerster S, Nevinny- Wagner A, Ewers R (2007) Comparison of laser surface scanning Stickel M, Lukas P (2001) A simple and non-invasive vacuum and fiducial marker-based registration in frameless stereotaxy. J mouthpiece-based head fixation system for high precision radio- Neurosurg 106(4):704–709 therapy. Strahlentherapie und Onkol 177(1):43–47 16. Sun Y, Luebbers H-T, Agbaje JO, Schepers S, Vrielinck L, 6. Martin A, Bale RJ, Vogele M, Gunkel AR, Thumfart WF, Lambrichts I, Politis C (2013) Validation of anatomical landmarks- Freysinger W (1998) Vogele–Bale–Hohner mouthpiece: reg- based registration for image-guided surgery: an in-vitro study. J istration device for frameless stereotactic surgery. Radiology Craniomaxillofac Surg 41(6):522–526 208(1):261–265 17. Hardy SM, Melroy C, White DR, Dubin M, Senior B (2006) A 7. Bardosi Z, Freysinger W (2016) Estimating FLE_image distribu- comparison of computer-aided surgery registration methods for tions of manual fiducial localization in CT images. Int J Comput endoscopic sinus surgery. Am J Rhinol 20(1):48–52 Assist Radiol Surg 11(6):1043–1049 18. Mascott CR, Sol J-C, Bousquet P, Lagarrigue J, Lazorthes Y, 8. Perwög M, Bardosi Z, Freysinger W (2018) Experimental vali- Lauwers-Cances V (2006) Quantification of true in vivo (applica- dation of predicted application accuracies for computer-assisted tion) accuracy in cranial image-guided surgery: influence of mode (CAS) intraoperative navigation with paired-point registration. Int of patient registration. Neurosurgery 59(1 Suppl 1):46–56 J Comput Assist Radiol Surg 13(3):425–441 19. Metzger MC, Rafii A, Holhweg-Majert B, Pham AM, Strong B 9. Fitzpatrick JM (2010) The role of registration in accurate surgical (2007) Comparison of 4 registration strategies for computer-aided guidance. Proc Inst Mech Eng H 224(5):607–622 maxillofacial surgery. Otolaryngol Neck Surg 137(1):93–99 10. Güler Ö, Perwög M, Kral F, Schwarm F, Bárdosi ZR, Göbel 20. Ma B, Moghari MH, Ellis RE, Abolmaesumi P (2010) Estima- G, Freysinger W (2013) Quantitative error analysis for computer tion of optimal fiducial target registration error in the presence of assisted navigation: a feasibility study. Med Phys 40(2):21910 heteroscedastic noise. IEEE Trans Med Imaging 29(3):708–723 11. Sonka M, Fitzpatrick JM (2000) Handbook of medical imaging, vol 21. Maurer CR, Maciunas RJ, Fitzpatrick JM (1998) Registration of 2. Medical image processing and analysis. SPIE The international head CT images to physical space using a weighted combination society for optical engineering, Bellingham of points and surfaces [image-guided surgery]. IEEE Trans Med 12. Fitzpatrick JM (2009) Fiducial registration error and target regis- Imaging 17(5):753–761 tration error are uncorrelated. In: Proceedings of SPIE, vol 7261 22. Kim S, Kazanzides P (2017) Fiducial-based registration with 13. Luebbers H-T, Messmer P, Obwegeser JA, Zwahlen RA, Kikinis a touchable region model. Int J Comput Assist Radiol Surg R, Graetz KW, Matthews F (2008) Comparison of different reg- 12(2):277–289 istration methods for surgical navigation in cranio-maxillofacial 23. Jeleff O (2005) Mikroskop- und pointerbasierte Navigation an der surgery. J Cranio-Maxillofac Surg 36(2):109–116 Laterobasis, MD-Thesis, Medizinische Universität Innsbruck 14. Woerdeman PA, Willems PWA, Noordmans HJ, Tulleken CAF, van der Sprenkel JWB (2007) Application accuracy in frameless image-guided neurosurgery: a comparison study of three patient- to-image registration methods. J Neurosurg 106(6):1012–1016

Journal

International Journal of Computer Assisted Radiology and SurgerySpringer Journals

Published: Jun 5, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off