Prostate SBRT on CyberKnife compresses what used to be 35-40 conventional fractions into just 5 high-dose treatments, each delivering 6.5-7.25 Gy. That kind of dose escalation only works if the beam is actually hitting the target, fraction after fraction, down to sub-millimetre precision. The problem is that the prostate doesn't hold still - it is bounded by a bladder that fills, a rectum that can carry gas, and it shifts in response to both, sometimes mid-fraction.
Unlike inter-fraction motion (the prostate being in a slightly different place from one treatment day to the next, which a pre-treatment CT or cone-beam scan can catch and correct), intrafraction motion happens during beam delivery itself - after setup has already been verified. If it goes undetected, the consequence is straightforward and unforgiving: the high-dose volume can under-cover the tumour, or spill extra dose into the bladder and rectum, the two organs-at-risk sitting closest to it.
CyberKnife is a robotic radiosurgery platform built specifically to answer that question in real time. Instead of a fixed gantry rotating around a couch, a linear accelerator is mounted on a six-degrees-of-freedom robotic arm, capable of firing non-isocentric, non-coplanar beams from hundreds of possible angles. Two orthogonal, ceiling-mounted diagnostic X-ray tubes paired with floor-mounted flat-panel detectors continuously image the patient during treatment, and the system's Target Localizing System (TLS) uses that imaging to track the tumour position and automatically steer the robotic arm to follow it.
For a soft-tissue target like the prostate - which isn't visible enough on its own with planar X-ray - CyberKnife relies on fiducial tracking. Three to five small gold markers are implanted directly into the gland under transrectal ultrasound guidance, roughly a week before planning, to allow tissue to stabilise around them and reduce the risk of migration. The markers are deliberately placed in a non-collinear, spread-out, triangular arrangement; if they end up roughly in a line, the system can't reliably resolve rotation, and tracking accuracy breaks down.
Six degrees of freedom: three translational axes (X - lateral, Y - longitudinal, Z - vertical, all in mm) and three rotational axes (roll, pitch, yaw, all in degrees).
In this study, a positional deviation of 2mm (translation) or 1° (rotation) triggered an automatic robotic correction to re-align the beam before the next pulse.
This is the mechanism the study is actually testing - not whether the prostate moves (it does, inevitably), but whether that movement stays inside the geometry CyberKnife's tracking system was built to handle.
This was a retrospective observational study conducted at the Department of Radiation Oncology, Amrita Institute of Medical Sciences and Research Centre, Kochi. It reviewed patients with histologically confirmed prostate adenocarcinoma who had completed CyberKnife SBRT using fiducial-based tracking, pulling the intrafraction displacement logs the system records automatically during every treatment.
Histologically confirmed prostate adenocarcinoma; treated with fiducial-tracked CyberKnife SBRT; 3+ implanted gold markers in acceptable (non-collinear) geometry; at least one completed fraction with usable tracking logs; age ≥18.
Non-fiducial tracking methods; fewer than 3 fiducials or poor marker geometry; significant fiducial migration or loss; incomplete tracking logs; prior pelvic radiation or prostatectomy; severe imaging artefacts.
Patients were positioned supine with a comfortably full bladder and empty rectum to keep anatomy reproducible between simulation and treatment. A 1.25mm-slice planning CT was fused with pelvic MRI for target delineation, and the Planning Target Volume was generated with a uniform 2mm margin around the Clinical Target Volume. Treatment plans were built on the Accuray Precision Treatment Planning System, using non-isocentric, non-coplanar beams to deliver 35-36.25 Gy in 5 fractions, prescribed to the 87-95% isodose line while respecting dose constraints to the bladder, rectum and femoral heads.
During each fraction, the TLS continuously acquired stereoscopic X-ray images, tracking fiducial position and logging translational (X, Y, Z) and rotational (roll, pitch, yaw) displacement at fixed intervals. For each of the 27 patients, across all 5 fractions, the study computed the mean, maximum, minimum and standard deviation of displacement along each of the six axes - 135 fraction-level records in total (27 patients × 5 fractions), rolled up into per-patient and overall summary statistics.
27 patients with histologically confirmed prostate adenocarcinoma were included, all treated to 36.25 Gy in 5 fractions. Across every fraction and every patient, the study found that intrafraction prostate motion was consistently small - and, critically, consistently within the range CyberKnife's tracking and correction system is designed to handle.
| Axis | Mean Shift | Std. Dev. | Minimum | Maximum |
|---|---|---|---|---|
| X - Lateral (mm) | 1.2 | 1.0 | 0.2 | 2.6 |
| Y - Longitudinal (mm) | 1.3 | 0.8 | 0.5 | 2.4 |
| Z - Vertical (mm) | 1.1 | 0.9 | 0.2 | 2.4 |
| Roll (degree) | 0.8 | 0.8 | 0.0 | 2.1 |
| Pitch (degree) | 1.5 | 1.3 | 0.2 | 3.1 |
| Yaw (degree) | 1.1 | 0.6 | 0.3 | 2.3 |
Values rolled up from per-fraction standard deviation, minimum, maximum and average displacement tables computed across all 27 patients and all 5 treatment fractions.
Pitch (rotation around the lateral axis - effectively the prostate tilting forward or backward) consistently showed the largest average displacement of the six parameters, at 1.5° mean and up to 3.1° maximum. That tracks with the anatomy: pitch is the rotation most sensitive to anterior-posterior changes in bladder and rectal filling, the two physiological factors that move the prostate the most from one moment to the next. Roll, by contrast, showed the smallest minimum shift - likely because there isn't much lateral soft-tissue structure next to the prostate to push it sideways, and external immobilisation (a Vac-Lok cradle) limits roll further.
What the data does not show is a drift over the course of treatment. Comparing fraction 1 against fraction 5 across all patients, standard deviation, minimum, maximum and mean values stayed essentially flat. There's no evidence that patients settle into worse positioning, or move more, later in a longer fraction - a reasonable concern with anxiety or fatigue, but not one borne out in this dataset.
Beyond the per-fraction summary statistics, the thesis includes a real-time correction trace from an actual treatment session: translational and rotational corrections plotted continuously over a 25-minute fraction.
This is really the practical version of the summary table above - it's one thing to know the mean shift was 1.5° in pitch across 27 patients; it's another to see, minute by minute, the robotic arm catching and correcting that drift in real time rather than letting it accumulate across a 25-minute delivery.
It's worth stepping back to why this level of tracking matters for prostate specifically. A conventional LINAC delivering IMRT or VMAT typically treats prostate cancer over 20-40 sessions at 1.8-2 Gy per fraction, correcting position between fractions (inter-fraction) using periodic imaging, but without continuous intrafraction tracking in most standard setups. CyberKnife compresses that into 4-5 sessions at a much higher dose per fraction - which only becomes clinically acceptable because of the real-time tracking this study is measuring.
| Feature | Conventional LINAC (IMRT/VMAT) | CyberKnife |
|---|---|---|
| Treatment Duration | 20-40 sessions | 4-5 sessions |
| Precision | High | Very high (real-time tracking) |
| Side Effects | Moderate | Lower (less exposure to healthy tissue) |
| Cost & Availability | Widely available, lower cost | Limited availability, higher cost |
None of this makes conventional LINAC treatment inferior - it remains the right, well-proven choice at most centres. But it does explain why a study quantifying intrafraction motion on CyberKnife matters beyond this one department: hypofractionated, high-dose-per-fraction SBRT is only as safe as the tracking system backing it up, and this is what that tracking actually looks like, measured in real patients rather than assumed from the equipment specifications.
The thesis closes with five conclusions, which I've kept close to the original wording:
Put simply: on this system, with fiducial tracking and a 2mm/1° correction threshold, the prostate moves - but not more than the technology built to follow it.
To be unambiguous about authorship: I did not design this study or carry out its analysis - that is Vidhya's work throughout. I did assist with data collection for it, as part of the clinical team at AIMS Kochi, and it's presented here as a summary for a broader clinical and technical audience.
The thesis was also certified by Dr. Debnarayana Dutta (Professor & HOD, Department of Radiation Oncology, AIMS Kochi) and Dr. Gireesh Kumar K.P (Principal, Amrita School of Medicine, AIMS Kochi), with the original acknowledgements additionally naming Dr. Srikanth Moorthy, Mrs. Parvathy Thejus and Ms. Ashitha M K for their support of the study.