Analysis of intrafraction motion in CyberKnife-based stereotaxy using mask based immobilization and 6D-skull tracking

Abstract

Purpose

Analysis of intrafraction motion in patients with intracranial targets treated with frameless, mask based stereotactic radiosurgery / radiotherapy using standard couch and 6D-skull tracking on CyberKnife.

Materials and methods

Twenty-seven treatment datasets of fifteen patients were analyzed. For each sequential pair of images, the correction to the target position (position “offset”) in six-degrees of motion was obtained. These offsets were used to calculate intrafraction shifts, and their statistical distribution. PTV margins were calculated, based on Van Herk formula.

Results

The mean ± 1 SD intrafraction translationals were 0.27±0.61mm in left-right, 0.24±0.62mm in antero-posterior and 0.14±0.24mm in supero-inferior direction, and rotations were 0.13±0.21 degrees roll, 0.18±0.25 degrees pitch and 0.28±0.44 degrees yaw. Most intrafraction shifts were ≤ 1mm and 1 degree. Fourteen instances of intrafraction shifts exceeding the robotic correction threshold were noted. Calculated PTV margins were 1mm, 1mm and 0.4mm for for left-right, antero-posterior and supero-inferior directions, respectively.

Conclusions

CyberKnife 6D-skull tracking with 1mm PTV margin effectively compensates for intrafraction motion. The occasional large intrafraction movements may assume significance for techniques not employing intrafraction motion management.

1. INTRODUCTION

Cranial Stereotactic radiosurgery (SRS) and radiotherapy (SRT) are techniques that deliver radiation in high dose per fraction with a high degree of conformality. The prescribed isodose line shapes conformally around the Planning Target Volume (PTV), with a region of steep dose fall-off in the periphery. Because of this sharp gradient between the target edge and normal tissue, treatment delivery necessitates a high level of precision to optimize tumour control and minimize normal tissue toxicity. The same requires strict immobilization of the head, while ensuring patient comfort, and set-up reproducibility. An accuracy of within 1 mm, and overall geometric and dosimetric uncertainty within 2 % is recommended [1].

The comparable levels of interfraction reproducibility shown by frameless techniques coupled with image guidance, and the impracticality of using rigid frames for SRT, have led to increased use of the former for both SRS and SRT [2-5]. However, there is no way to limit the intrafraction motion with the frameless technique apart from stringent intrafraction imaging, and hence, the need to assess the effectiveness of the same is vital. It may be emphasized that uncorrected intrafraction motion may jeopardize the systematic efforts of radiation planning, leading to inferior target coverage and serious normal tissue toxicity.

Initial encouraging experience for intracranial lesions treated with SRS / SRT on Gamma Knife (GK) [6], and subsequently on Linear Accelerator (LINAC) [7,8] has paved the way for its execution on the highly advanced robotic radiation delivery system, the CyberKnife (CK). This machine uses a 6-MV LINAC mounted on a fully articulated robotic arm with six degrees of freedom, which allows targeting lesions from a much wider variety of angles than on a conventional LINAC [9]. Moreover, the processes of calculation of positional shifts and their correction, intrafraction imaging and tumour tracking are fully automated. The purpose of the present study is to demonstrate the magnitude of intrafraction motion in cranial lesions treated with thermoplastic mask based frameless SRS / SRT coupled with 6D-skull tracking on the CyberKnife® VSI Radiosurgery System (Accuray Inc., Sunnyvale, CA, USA). A secondary objective was to determine the adequate PTV margin that would be sufficient to compensate for this intrafraction motion.

2. MATERIALS AND METHODS

Fifteen patients treated with cranial SRS / SRT on CK using Skull tracking were identified as subjects for the study. Patients were immobilized in supine position using a customized, non-invasive uniframe thermoplastic cast. No external fiducial markers were used because of the capability of image based couch correction by CK at set up. A non-contrast CT scan of the head and a contrast enhanced T1weighted MRI, both with 1 mm slice thickness, were done and fused together for treatment planning. Gross tumor volume (GTV) was delineated as contrast enhancing tumor on MRI. Clinical Target Volume (CTV) and dose/fractionation were decided on a case-by-case basis. 1 mm PTV was given. SRS was selected to treat tumors smaller than 3 cm, while larger lesions were treated with SRT. Inverse planning was done using the MultiPlan software (Accuray Inc., Sunnyvale, CA, USA). An optimal treatment plan was expected to deliver SRS / SRT to the 80%-90% isodose line encompassing 100 % of the PTV. Target coverage, dose heterogeneity and conformality index were examined to evaluate the quality of treatment plans. Strict quality assurance was done to ensure smooth execution of the treatment plan on the machine.

For treatment, the patient with the thermoplastic cast assembly was positioned on the treatment couch in an identical manner as per CT simulation, using in-room lasers. Two in-room ceiling mounted kilovoltage (kV) X-Ray sources placed at 45 degress to vertical along with flat panel detectors provide real-time imaging of the bony anatomy of the patient. The pretreatment images are coregistered with the planning digitally reconstructed radiographs (DRRs) using the two-dimensional (2D)-three-dimensional (3D) image registration method, as described by Fu et al [9], and necessary set up errors are detected and corrected by couch movement, thus bringing the patient in treatment position. After the treatment starts, real time set up errors are continuously recorded and adequate adjustments done by the robotic arm, without moving the patient. The robot can correct translations of up to ±10 mm and rotations of up to ±1.0◦ roll and pitch, and ±3.0◦ yaw [10], with a system targeting error of ≤ 0.95 mm Root Mean Squared (RMS) [11]. A lesser correction threshold is optional, and we set the same at 5 mm for translations, 1 degree for roll and pitch, and 3 degrees for yaw. When acquisition starts, a default image interval is selected. After verifying target stability for the first few minutes, the imaging interval can be adjusted to a value between 5 to 150 seconds, with due adjustments according to the target position. During treatment, each projection is co-registered with the reference DRR image in six-dimensional space, thus providing data on intrafraction motion, and its simultaneous correction by the robot. In case an offset greater than the robotic threshold of correction is encountered, the treatment is automatically interrupted, and couch movement is required to correct the same. The clinical characteristics of patients pertaining to the dataset are detailed in Table 1.

Table 1.

Patient Characteristics

Sl No	Age/Sex	Diagnosis	Radiation dose schedule
1	63/F	Meningioma, post op, with residual	18 Gy in 3 fractions
2	69/F	Cerebellopontine angle Schwannoma	18 Gy in 3 fractions
3	28/M	Cerebellar A-V malformation	22 Gy in 1 fraction
4	43/M	Pituitary Adenoma, post op, with residual	18 Gy in 3 fractions
5	55/F	Cerebellopontine angle Schwannoma	11.5 Gy in 1 fraction
6	64/M	Meningioma, post op, with residual	18 Gy in 3 fractions
7	22/M	Frontal A-V malformation	20 Gy in 1 fraction
8	50/M	Cerebellopontine angle Schwannoma	18 Gy in 3 fractions
9	28/F	Cerebellopontine angle Schwannoma	16.5 Gy in 3 fractions
10	68/F	Carcinoma Breast with single brain metastasis	20 Gy in 1 fraction
11	49/F	Carcinoma Breast with single brain metastasis	20 Gy in 1 fraction
12	25/F	Parietal A-V malformation	15 Gy in 1 fraction
13	24/F	Recurrent Temporal A-V malformation	18 Gy in 1 fraction
14	19/M	Frontal A-V malformation	20 Gy in 1 fraction
15	60/M	Carcinoma Lung with single brain metastasis	22 Gy in 1 fraction

A-V malformation = 5

Cerebellopontine angle Schwannoma = 4

Brain metastasis = 3

Meningioma = 2

Pituitary Adenoma = 1

The complete details of the positional offset values and the consequent robotic movements required to correct them during treatment are recorded as system log files. The individual patient dataset consists of translational offsets in lateral or left-right (LR) direction, antero-posterior (AP) direction and supero-inferior (SI) direction, and rotational offsets as roll, pitch and yaw. These datasets were screened for offsets due to manual couch movement during set-up phase or in-between treatment, and these were excluded from the analysis. The actual intrafraction shifts in each direction were calculated by subtracting a particular offset value (n) from the immediately succeeding one (n + 1), as follows:

Translations:

DX₁ = X_n+1 – X_n DY₁ = Y_n+1 – Y_n DZ₁ = Z_n+1 – Z_n

DX₂ = X_n+2 – X_n+1 DY₂ = Y_n+2 – Y_n+1 DZ₂ = Z_n+2 – Z_n+1

DX₃ = X_n+3 – X_n+2 DY₃ = Y_n+3 – Y_n+2 DZ₃ = Z_n+3 – Z_n+2 …. and so on

Rotations:

DR₁ = R_n+1 – R_n DP₁ = P_n+1 – P_n DYW₁ = YW_n+1 – YW_n

DR₂ = R_n+2 – R_n+1 DP₂ = P_n+2 – P_n+1 DYW₂ = YW_n+2 – YW_n+1

DR₃ = R_n+3 – R_n+2 DP₃ = P_n+3 – P_n+2 DYW₃ = YW_n+3 – YW_n+2 ... and so on

where

DX₁ is the first intrafraction shift in the Left-Right (LR) direction, X_n+1 and X_n are the second and first positional offsets, respectively, in the LR direction

DY₁ is the first intrafraction shift in the Antero-Posterior (AP) direction, Y_n+1 and Y_n are the second and first positional offsets, respectively, in the AP direction

DZ₁ is the first intrafraction shift in the X Supero-inferior (SI) direction, Z_n+1 and Z_n are the second and first positional offsets, respectively, in the SI direction

DR₁ is the first intrafraction shift in Roll, R_n+1 and R_n are the second and first positional offsets, respectively, in the Roll

DP₁ is the first intrafraction shift in Pitch, P_n+1 and P_n are the second and first positional offsets, respectively, in the Pitch

DYW₁ is the first intrafraction shift in Yaw, YW_n+1 and YW_n are the second and first positional offsets, respectively, in the Yaw

For calculation of mean shifts, the DX to DYW values were converted to “all-positive” values by eliminating the minus sign. This was done so as to avoid the “+” and “–” values mutually cancelling out each other in the calculation. Mean intrafraction shifts and Standard Deviations were then calculated from this all positive value dataset for each translational and rotational direction, individually for each fraction and also as the pooled dataset of 27 fractions of fifteen patients. The 3D-vector was calculated as the Root Mean Square (RMS) value of the individual mean translational shifts. Mean 3D-vector was the mean of 3D-vectors of all 27 individual fraction datasets.

PTV Margin calculation was done based on the Van Herk formula [12]:

PTV margin = 2.5å + 0.7s

where å is the systematic error and s is the random error.

For determining the systematic error, mean shifts in a particular direction (LR / AP / SI) of individual treatment fractions were considered. The standard deviation (SD) of these mean shifts provided the value of å. For determining the random error, the SD of shifts in a particular direction (LR / AP / SI) of individual treatment fractions were considered. The RMS value of this SD provided the value of s. All statistical analyses were performed using Microsoft Office Excel version 2007.

3. RESULTS

Mean treatment time for daily fraction was 41 minutes (range 27 to 63 minutes). Mean imaging interval during delivery of a fraction was 36.5 seconds (range 20 to 45 seconds). The mean intrafraction translational shifts were 0.27 mm (SD ± 0.61 mm) in the L-R direction, 0.24 mm (SD ± 0.62 mm) in A-P direction, and 0.14 mm (SD ± 0.24 mm) in the S-I direction. Mean 3D-vector for translational shifts was 0.45 mm. In all, 96.9 % (2955/3048), 97.7 % (2979/3048), and 98.9 % (3016/3048) of the translational shifts were within 1, 2, and 3 mm, respectively. The rotational shifts were 0.13 degree roll (SD ± 0.21 degree), 0.18 degree pitch (SD ± 0.25 degree), and 0.28 degree yaw (SD ± 0.44 degree). 97.3 % (2965/3048), 99.5 % (3034/3048), and 100 % (3048/3048) of the rotations were within 1, 2, and 3 degrees, respectively. There were five instances when the robotic threshold of 5 mm translational correction was exceeded, necessitating treatment interruption and couch correction. There were nine such instances in case of rotational corrections. The individual mean intrafraction errors for the 27 fractions are listed in Table 2. The outliers are marked with bold italicized font. Figure 1 depicts an example of the intrafraction shifts for an individual patient, and Figures 2 and 3 depict the range of shifts in all directions for the 27 treatment fractions. The outlier fractions with especially large range of shifts are encircled. The calculated systematic error in the LR, AP and SI directions was 0.27mm, 0.24mm and 0.08mm, respectively and the random error was 0.56mm, 0.70mm and 0.22mm, respectively. PTV margin recipe, as per Van Herk calculations, was 1mm, 1mm and 0.4mm for LR, AP and SI directions, respectively.

Table 2.

Intrafraction Shifts of 27 SRS* / SRT^† fractions

Sl No	Mean L/R‡ shift ± SD§ (mm)	Mean A/P# shift ± SD§ (mm)	Mean S/I$ shift ± SD§ (mm)	Mean Roll ± SD§ (degrees)	Mean Pitch ± SD§ (degrees)	Mean Yaw ± SD§ (degrees)
1	0.37 ± 0.69	0.40 ± 1.13	0.08 ± 0.08	0.04 ± 0.06	0.13 ± 0.12	0.24 ± 0.37
2	0.27 ± 0.52	0.31 ± 0.69	0.12 ± 0.13	0.13 ± 0.16	0.28 ± 0.34	0.47 ± 0.59
3	0.27 ± 0.55	0.16 ± 0.21	0.13 ± 0.20	0.08 ± 0.08	0.16 ± 0.24	0.33 ± 0.49
4	0.16 ± 0.22	0.14 ± 0.19	0.08 ± 0.08	0.18 ± 0.36	0.08 ± 0.10	0.14 ± 0.20
5	0.34 ± 0.61	0.43 ± 0.71	0.15 ± 0.16	0.21 ± 0.27	0.33 ± 0.25	0.40 ± 0.51
6	0.11 ± 0.16	0.20 ± 0.19	0.14 ± 0.15	0.11 ± 0.10	0.17 ± 0.26	0.14 ± 0.16
7	0.27 ± 0.93	0.11 ± 0.13	0.07 ± 0.08	0.10 ± 0.18	0.11 ± 0.15	0.15 ± 0.22
8	0.12 ± 0.20	0.13 ± 0.15	0.12 ± 0.13	0.04 ± 0.06	0.09 ± 0.09	0.11 ± 0.15
9	0.15 ± 0.26	0.09 ± 0.12	0.11 ± 0.16	0.07 ± 0.15	0.13 ± 0.18	0.15 ± 0.15
10	0.23 ± 0.26	0.16 ± 0.18	0.11 ± 0.12	0.22 ± 0.28	0.24 ± 0.24	0.34 ± 0.44
11	0.41 ± 0.70	0.42 ± 0.76	0.11 ± 0.09	0.08 ± 0.09	0.25 ± 0.25	0.32 ± 0.35
12	0.10 ± 0.13	0.26 ± 0.77	0.12 ± 0.27	0.06 ± 0.07	0.10 ± 0.13	0.17 ± 0.25
13	0.20 ± 0.45	0.13 ± 0.25	0.15 ± 0.28	0.12 ± 0.12	0.15 ± 0.15	0.17 ± 0.14
14	0.35 ± 0.72	0.29 ± 0.59	0.14 ± 0.32	0.12 ± 0.18	0.17 ± 0.22	0.17± 0.18
15	0.17 ± 0.17	1.21 ± 1.27	0.26 ± 0.20	0.23 ± 0.17	0.34 ± 0.35	1.02 ± 1.14
16	1.41 ± 1.12	0.23 ± 0.16	0.49 ± 0.49	0.46 ± 0.34	0.57 ± 0.61	0.77 ± 0.55
17	0.45 ± 0.86	0.29 ± 0.71	0.12 ± 0.11	0.19 ± 0.24	0.21 ± 0.26	0.26 ± 0.43
18	0.13 ± 0.12	0.11 ± 0.14	0.05 ± 0.07	0.16 ± 0.27	0.11 ± 0.14	0.37 ± 0.70
19	0.17 ± 0.27	0.08 ± 0.15	0.11 ± 0.11	0.06 ± 0.07	0.13 ± 0.13	0.13 ± 0.13
20	0.14 ± 0.18	0.22 ± 0.41	0.13 ± 0.12	0.12 ± 0.23	0.13 ± 0.17	0.25 ± 0.35
21	0.07 ± 0.12	0.09 ± 0.09	0.19 ± 0.45	0.05 ± 0.07	0.13 ± 0.24	0.22 ± 0.59
22	0.26 ± 0.80	0.17 ± 0.20	0.13 ± 0.13	0.15 ± 0.21	0.13 ± 0.18	0.29 ± 0.55
23	0.07 ± 0.09	0.14 ± 0.16	0.09 ± 0.12	0.06 ± 0.11	0.12 ± 0.14	0.20 ± 0.19
24	0.17 ± 0.12	0.11 ± 0.12	0.07 ± 0.08	0.42 ± 0.38	0.11 ± 0.07	0.53 ± 0.64
25	0.14 ± 0.19	0.67 ± 1.22	0.19 ± 0.25	0.22 ± 0.32	0.21 ± 0.25	0.33 ± 0.43
26	0.20 ± 0.37	0.54 ± 1.33	0.21 ± 0.47	0.04 ± 0.05	0.11 ± 0.09	0.19 ± 0.30
27	0.70 ± 1.31	0.15 ± 0.11	0.14 ± 0.10	0.09 ± 0.06	0.13 ± 0.09	0.23 ± 0.18

^*SRS: Stereotactic Radiosurgery,

^†SRT: Stereotactic Radiotherapy

^‡L/R: Left-Right;

^#A/P: Antero-Posterior;

^$S/I: Supero-Inferior

^§SD: Standard Deviation

Outliers marked with bold italicized font

Figure 1.

Intrafraction translational and rotational shifts plotted against time, for a treatment fraction.

SRS: Stereotactic Radiosurgery; SRT: Stereotactic Radiotherapy

L/R: Left-Right, A/P: Antero-Posterior, S/I: Supero-Inferior

4. DISCUSSION

The current study demonstrated that the majority of the intrafraction motion was less than 1 mm translation (96.9 %) and 1 degree rotation (97.3 %) during treatment delivery on CyberKnife with mask based immobilization and 6-D skull tracking, with notable presence of occasional large shifts (> 5 mm translation, > 1º roll and pitch, > 3º yaw). Calculated PTV margins were 1 mm or less in the three translational directions.

Since the late sixties when the first cranial radiosurgery was done on Leksell’s Gamma Knife [6], different intracranial lesions have been treated with SRS / SRT with varying levels of accuracy. Improvements in radiation technology and physics have been ably supported by intelligent use of principles of radiation biology to attain the therapeutic aim [13,14]. Besides GK, SRS / SRT have been increasingly practiced on LINAC and more recently on CK. Owing to the facility of 6-D skull tracking on CyberKnife, minimal PTV margins are given for SRS and SRT. The clinical significance of reduced treatment margins in SRS has been evaluated by Nataf et al [15]. In their study, on enlarging the PTV from 1 to 2 mm, significantly more severe parenchymal complications occurred within a group of 93 patients with brain metastases. They recommended reduction of the margin to 1 mm for minimization of complications.

With minimal margin for error, accurate and reproducible immobilization of head becomes fundamental to the delivery of SRS and SRT. Though the same required a rigid cranial frame in the past, now diverse frameless devices are available with comparable accuracy as frame based techniques for minimizing interfraction movement [16]. Based on the results in literature [2-4] and that of our own LINAC study [5], we believe that a well constructed thermoplastic mask with dedicated image guidance in real time during treatment serves the purpose equally well. In our LINAC study [5], the mean translational set-up errors with framebased SRS were 1 mm (SD=0.3) in LR direction, 0.2 mm (SD = 1.2) in SI direction, and 0.1 mm (SD = 0.3) in AP direction, and rotational errors were: roll 0.3 degree (SD = 0.7), pitch 0.4 degree (SD = 0.7), and yaw 0.2 degree (SD = 0.4). For frameless SRS, the set-up errors were 0.4 mm (SD = 0.9) in LR direction, 1.1 mm (SD = 1.1) in SI direction and 0.5 mm (SD = 1.3) in AP direction, and rotational errors were: Roll 0.1 degree (SD = 0.8), pitch 0.2 degree (SD = 0.4), and yaw 0.3 degree (SD = 0.4). Despite the comparability of the two techniques regarding set-up errors, the same, however, may not be true in terms of minimizing intrafraction motion.

Till date, most studies assessing intrafraction motion have been performed for LINAC based radiosurgery, relying on pre- and post-treatment imaging as surrogate. Badakhshi, et al [17] performed measurements at several time points during the course of treatment and found that 12% of the intrafraction values in the three dimensions were above the safety margin of 1 mm typically applied in SRS. Murphy et al [18] examined patterns of patient movement during frameless image-guided radiosurgery and found that the mean translational difference was 0.45 mm per axis. Inoue, et al [19] reported median translational residual patient motion as 0.1 mm for each axis, and rotational residual patient motion as 0.1 degree for pitch and roll and 0.2 degree for yaw. Due to this intrafraction motion, the consequent dose error for D95 was reported as within 1 % in more than 95 % of cases, and the maximum dose error for D10 to D90 was within 2 %. Kang et al [20] examined the dosimetric impact of intrafraction movements occurring during image-guided frameless brain radiosurgery and attempted to derive optimal margins required to compensate the movement. They derived the formula 1.0 r + 0.2s, where r and s are the average and standard deviation of the movements, respectively. They calculated that the optimal margins for treatment times of 10, 20, and 30 min were 2.1, 3.2, and 4.2 mm, respectively, at 90% confidence level. They, however, did not consider the impact or significance of correcting these intra-fraction errors on the dosimetry, which, we believe, would have resulted in a much smaller margin recipe. Intrafraction target shift, and the accompanying dosimetric uncertainty, has thus been shown to be a significant event in literature.

Murphy [21] and Hoogeman [22] assessed intrafraction motion on CK, and calculated margins based on the relation proposed by Van Herk et al [12]. Murphy et al [21] showed that if periodic intra-fraction alignment corrections are made, the effect of intra-fraction shift is mostly eliminated and the margin is reduced to 1.2 mm for cranial cases, with systematic error as 0.1mm and random error as 1.2 mm. Hoogeman et al [22] showed that the SD of the systematic intrafraction displacements for a 15 minute intracranial treatment fraction was 0.8 mm, and increased linearly with time. Floriano, et al [23] reported their experience of cranial SRS / SRT on CK with a 99% displacement error less than 0.85 mm, systematic intrafraction movement components less than 0.05 mm and random intrafraction movement components less than 0.3 mm in the 3 translational axes.

Figure 2.

Range of Intrafraction translational shifts for the 27 treatment fractions

SRS: Stereotactic Radiosurgery; SRT: Stereotactic Radiotherapy

L/R: Left-Right, A/P: Antero-Posterior, S/I: Supero-Inferior

In our study, the observed mean translational and rotational shifts were all less than 0.3 mm and 0.3 degrees, with > 97 % of shifts less than 1 mm and 1 degree. These are comparable to the shifts obtained in the study by Inoue et al [19], and much lesser than some of the earlier CK studies mentioned above [21-23] As explained by Inoue et al [19], this difference might be attributable to the interval between acquiring sequential images and the type of imaging system used in each study. The mean interval between sequential images was 36.5 s in our study, which was much lesser than the mean interval of 120 s in the study by Murphy et al [21]. Improved imaging quality provided by the amorphous silicon diode detectors, over and above the X-ray image intensifiers used earlier, also accounts for the lesser intrafraction shifts.

Figure 3.

Range of Intrafraction rotational shifts for the 27 treatment fractions

SRS: Stereotactic Radiosurgery; SRT: Stereotactic Radiotherapy

The systematic error of not more than 0.3 mm, random error of not more than 0.7 mm, and the calculated PTV margin of 1 mm and less, based on translational shifts, as seen in our study, is in agreement with earlier studies [21-23], and concurs with the recommended PTV margins for cranial SRS [1]. As per Guckenberger et al [24], a 0.6 mm translational margin is sufficient to correct for a 1º rotational misalignment. Considering the effect of rotation in addition to translations on overall margins, and that mean rotational shifts being less than 0.3 degrees in our study, it appears that a 1mm translational margin would most certainly correct for rotations as well. All this put together, and the fact that CK corrects all the intrafraction movements, leads us to conclude that a 1 mm PTV margin is sufficient for intracranial targets treated by 6-D skull tracking.

A word of caution, however, needs to be added in view of the observation of occasional large intrafraction shifts exceeding 5 mm. The frequent intrafraction imaging inherent in the treatment workflow of CK allows us to detect the presence of such a movement, something which goes undetected in systems not employing any form of intrafraction imaging. Our mean frequency of imaging was 36.5 seconds, which is still not real-time in the strict sense, and it is the targeting error between two successive image acquisitions that is used as surrogate for intrafraction motion. Thus, CK may still fail to detect and evaluate the impact of real time transient motion, which, however, must be small for intracranial targets. Yet, the treating physician should be aware that intrafraction motion ≥ 5 mm does occur occasionally and hence the same may be accounted for in the PTV margin, especially where intrafraction imaging is not being used, and considering that systematic uncertainties (e.g. in delineation) are also routinely unaccounted for.

The mean intrafraction shift values recorded were specific to the 5-mm translational, 1 degree roll and pitch, and 3 degree yaw threshold used in tracking because the greater-than-threshold shifts were reset by couch adjustments, and hence, per se, were not a part of intrafraction motion calculations. The mean shifts therefore do not exactly represent the mean magnitude of actual intrafraction motion during a complete treatment fraction on CyberKnife. However, the difference from the actual mean is likely to be very small (< 0.1 mm).

There are reports of a recent image projection technique called augmented reality-guided neurosurgery, with a reported projection error of 0.8 ± 0.25 mm [25]. Other techniques of neuro-navigation described in literature attain accuracy in the range of 1.5 mm [26,27]. The CyberKnife thus offers a solution to cure a large number of intracranial lesions non-invasively, without the need for anaesthesia, and with a greater accuracy than conventional neuro-navigation techniques. Its increased use in treating benign and malignant cranial lesions is therefore recommended.

5. CONCLUSION

In our experience, intrafraction motion for intracranial targets treated with fiducial free, frameless cranial SRS / SRT on CyberKnife with 6-D skull tracking, is within the acceptable range, and can be reliably detected and corrected. A PTV margin of 1 mm appears adequate to account for most of the intrafraction motion in this situation. However, significant intrafraction motion occurs during treatment delivery when mask based immobilization is used, and hence the same should be accounted for, in situations where intrafraction imaging is not being practiced. Owing to the highest level of precision, excellent automation, ease of treatment planning and delivery and avoidance of anaesthesia, CyberKnife Stereotactic Radiosurgery / Radiotherapy is highly recommended as an alternative to complex cranial neurosurgical procedures.