The Development of Technology for Effective Respiratory-Gated Irradiation Using an Image-Guided Small Animal Irradiator

The recent development of image-guided small animal irradiators represents a significant improvement over standard systems. These devices enable preclinical studies to mimic radiotherapy in humans, and have great potential for facilitating the translation of radiobiological research into the clinic (1, 2 ). Furthermore, these irradiators are able to utilize a cone beam CT (CBCT) reconstructed image to guide the delivery of tightly collimated targeted beams in conjunction with gantry or couch rotation, allowing for the required dose delivery to the tumor while sparing normal tissue. Two commercial systems are currently available: the Xstrahl, Small Animal Radiation Research Platform (SARRP), developed in collaboration with Johns Hopkins University (3, 4 ) with treatment planning using superposition convolution techniques (Muriplan from Xstrahl Life Sciences) (5 ), and the Precision X-ray, X-RAD 225Cx, developed in conjunction with the Princess Margaret Hospital (Toronto, Canada) (6 ) with treatment planning available using Monte Carlo methods (SmART-Plan from Precision X-ray, Inc.) (7 ). In addition, a number of other noncommercial small animal radiation research platforms have been developed to allow precise irradiation of structures in small animals (8–11 ) along with a range of treatment planning solutions (12 ).

We describe here a new treatment head assembly to further facilitate the irradiation of small animals. In particular, the development of a fast X-ray shutter, along with a noncontact, optical method for monitoring breathing, allows the delivery of respiratory gated collimated beams during the period when tissue movement is minimal. Respiratory motion during treatment is a concern due to the movement of organs and tumor as a function of time, especially those in close proximity to the diaphragm. As a result, there is the potential for the tumor (or targeted region) to partially move out of the treatment field for some period of time, thus reducing average dose and dose heterogeneity to the target volume. The development and implementation of respiratory gated beam delivery will enable the improved precision of these irradiators to be fully exploited within the thorax and abdomen, helping to ensure optimum dose delivery to the target area while minimizing dose to surrounding normal tissue. These developments follow those in clinical practice where respiratory gated radiotherapy techniques are being implemented to manage target and organ motion during treatment (13 ) and represent an important advance for preclinical radiotherapy studies (14 ).

Imaging and irradiations were performed using an Xstrahl Small Animal Radiation Research Platform (SARRP) irradiator (Camberley, UK) (3, 4 ). This consists of a constant voltage, dual focus 225 kV_P X-ray source (Varian NDI-225-22) mounted on a rotating arm, robotic stages on which the mouse is mounted and positioned, providing four degrees of freedom. A flat panel digital X-ray detector (XRD 0829 AN14; PerkinElmer®, Waltham, MA) was used for acquiring CBCT projections. Typically, the tube is operated at 100 kV_P, 0.3 mA using the fine focus (EN12543: 1.0 mm) and a 1.0-mm aluminium filter for imaging and 220 kV_P, 13 mA using the broad focus (EN12543: 3.0 mm) and a 0.15-mm copper filter for delivery of treatment beams. The existing collimator section supplied by the manufacturer has been upgraded by replacement of the treatment head with an in-house head assembly (see Fig. 1) and incorporates the following:

• •

  a fast and silent X-ray shutter to control beam delivery and allow respiratory gated irradiations;

• •

  a motorized X-ray filter slider for automated selection of imaging and treatment beam hardening filters;

• •

  a transmission ionization chamber, locally corrected for temperature and pressure variations, to monitor dose delivery;

• •

  laser illumination for checking beam position and manual alignment;

• •

  a kinematic location system enabling easy removal and replacement of the collimator assembly in a reproducible fashion; and

• •

  bespoke collimators.

The laser illumination and the collimators are completely removable from the rest of the assembly (during imaging) but are normally held together (during irradiation) with several pairs of strong magnets, with electrical connections provided by spring contacts. The location system between the two is based on a modified Maxwell type of kinematic design (15 ), with each of the “V”s replaced by two parallel cylinders that locate three spheres. The parallel cylinders point towards the beam center. Such an arrangement permits accurate and highly reproducible methods of separating and relocating these two subsystems.

The X-ray shutter, shown in Fig. 2, is placed directly behind the X-ray tube output port, made of a 16-mm diameter lead cylinder encased in an 18-mm diameter stainless-steel holder. A 10 mm × 22.75 mm aperture is machined through the holder and the lead. When the central axis of the beam is aligned to the aperture, the X ray along the central axis is shielded using lead (6 mm) and stainless steel (2 mm) materials, resulting in a measured transmission <0.02% for a 220 kVp copper-filtered beam with a half-value layer (HVL) of 0.79 mm of copper, as shown in Fig. 2C. When the aperture is aligned with the beam, the resulting X-ray beam “fan” is large enough to illuminate the entire mouse. Subsequent components in the beam path are wide enough to accommodate this “fan”, as is normally used for CBCT imaging, when the collimator assembly is removed.

The shutter cylinder is driven by a proportional rotary solenoid (GDR; Magnet Schultz Ltd, Old Woking, UK), which incorporates an analogue rotation angle sensor. This sensor is used as part of a proportional-integral-derivative type of feedback loop to continuously monitor and control the cylinder's angular position (as shown in Fig. 3). The shutter solenoid, rated at a nominal 12 V, is significantly overdriven, up to 100 V peak-to-peak at the start of the motion, to speed up transitions and overcome the rate-limiting effects of the solenoid's inductance. This ensures fast operating times while the feedback loop accurately controls the final positions. The solenoid is driven using a pair of pulse-width-modulated drivers operating at ∼200 kHz and used in an H-bridge type of arrangement. The H-bridge switches (Q1–Q4) are driven such that either switches Q1, Q4 or switches Q2, Q3 are closed at any one time and provide bidirectional drive to the solenoid. The switches are galvanically isolated from the rest of the electronics and screened to prevent inevitable, but unwanted interference. Diodes D ensure that the P-channel (Q1, Q3) and the N-channel (Q2, Q4) are protected from the solenoid's back-EMF voltage while potential shoot-through is eliminated by applying ∼50 ns “all switches off” delays in between transitions of the pulse-width modulated drive waveform. This arrangement is very efficient, essentially providing power only during the acceleration and deceleration phases of the gating process and thus maintaining low driver and solenoid operating temperatures. The shutter is quite capable of following even a 10 Hz command signal, well in excess of the animal's breathing frequency.

Some latency is inevitable in any mechanical system, but this does not exceed 20 ms, as shown in Fig. 4. Nevertheless, since a collimated beam is passed through a much narrower range of angles than the 90° shutter rotation, as dictated by the X-ray-beam collimator-defined diameter, both the latency and the turn-on/off times are significantly reduced, to <12 ms and <3 ms, respectively, for a 3-mm diameter X-ray beam. The shutter has been demonstrated to be capable of working effectively at frequencies up to 5 Hz (see Supplementary Fig. S1; http://dx.doi.org/10.1667/RR14753.1.S1), which corresponds to a respiratory rate of 300 breaths per minute (BPM).

The solenoid can rotate up to 110°, nonetheless, we only use a 90° rotation, centered within the 110° operating range. The Hall-effect noncontact angular position sensor fitted to the solenoid provides angular feedback information, and thus a shutter position signal. A positional feedback arrangement ensures that the solenoid's end-stops are never reached. This ensures silent operation, an important consideration when this system is used with lightly anesthetized mice that are liable to twitch in response to audible high-frequency sounds.

For therapeutic exposures, the shutter can be opened for either a predetermined time (calculated from the dose rate) or for a given number of monitor units (MU) using the transmission ionization chamber, described below. More importantly, the shutter can also be controlled using a signal from the adaptive respiratory-gated system, described in section: “Adaptive Respiratory Gating”. This restricts dose delivery to the period during the breathing cycle when tissue movement is minimal.

An X-ray filter slide assembly, allowing for up to five positions for beam hardening filters and/or apertures, is included in the X-ray beam path following the shutter. This assembly is removable and motorized, and enables automated selection of imaging and treatment filters as well as potentially useful beam restriction apertures.

The removable slide assembly is driven using a DC motor, rotary encoder and gearbox followed by a rack-and-pinion arrangement. The slide assembly is fitted with two small magnets and a single, centrally mounted, Hall-effect position sensor. During initialization, the number of motor-encoder pulses required to travel between the magnets is determined while the geometry of the slider positions (and the magnets) defines the absolute position of the slider. After this initialization procedure, the drive system calculates the number of motor encoder pulses required between the current and the target filter positions. The motor encoder provides quadrature-type outputs such that the position and the direction of travel are known. Furthermore, the target position is reached using a zero-hysteresis method, whereby the target position is always reached in the same travel direction. Finally, a few millimeters before the target position is reached, the motor slows down to eliminate overrun due to system inertia. Although the feedback signal in this system is obtained indirectly, by sensing motor shaft rotation rather than slider position, the accuracy and repeatability remain within 0.25 mm, which is more than adequate for this type of application.

A transmission type of monitor ionization chamber, shown in Fig. 5, follows the shutter and a beam filtration arrangement. The atmospheric air-filled monitor chamber is constructed from two 0.5-mm thick FR4 (a composite material composed of woven fiberglass cloth with an epoxy resin binder) printed circuit boards. One of these supports a polarizing electrode, while the other supports a guarded collection electrode and a screening electrode on the other side of the board; both of these are 25-μm-thick copper layers, i.e., the X rays pass through 75 μm of copper and 1 mm of FR4 material in total. The 6.2-mm thick air-filled chamber is operated at a bias voltage of –600 V, ensuring collection of all charge in the chamber. The high voltage is generated from a flyback transformer type of DC-DC converter.

The chamber current is processed by a programmable gain current-to-voltage converter (I-V) followed by a multiplying type of digital-to-analogue converter (MDAC, 16-bit resolution). The I-V converter gain can be increased to allow more sensitive dose measurements to be performed, such as during CBCT imaging. The MDAC is supplied with a digital value inversely proportional to the chamber atmospheric pressure (in mmHg) and proportional to chamber temperature (in K), normalizing the chamber response to STP (273.15 K/760 mm pressure) air density. At this point a compensated dose-rate signal is available for quality assurance purposes. This signal is then integrated with a voltage-to-frequency converter (maximum frequency 100 kHz) followed by a counter to provide a count proportional to delivered dose. The voltage-to-frequency pulses are counted in a 32-bit counter preceded by an 8-bit prescaler. This arrangement acts as a drift-free, very wide dynamic range “dose” monitoring system, compensated for chamber temperature and pressure, providing continuously compensated dose and dose rate information.

In addition to serving as a quantitative check of X-ray delivery during irradiations, different calibration factors for the various collimators can be used at the MDAC input to set the integrator output to 1 cGy/MU calibration, as is commonly used in clinical radiotherapy. This then provides the user the option for beam delivery to be defined in MUs rather than irradiation time. The “shutter open/close” feedback, shown in Fig. 3, can also be used to gate a clock (100 μs period) and count the resulting pulses to determine true irradiation time. This provides a useful check when beam delivery is defined in MUs, or it can be used to define an irradiation in terms of time rather than MUs. Clearly, a totalized time measurement (or setpoint) will be limited in accuracy by inevitable small differences in shutter off-to-on and on-to-off transition times as well as by the number of shutter operations during the exposure. This makes the use of the ionization chamber setpoint always preferable for accurate delivered dose monitoring. All the counting, control and monitoring functions are performed locally with a PIC®-type microprocessor (Microchip Technology Inc., Chandler, AZ) using an I2C (http://www.i2c-bus.org/i2c-bus/) interface, operating at 100 kb/s, to an off-chamber assembly USB interface.

The new Oxford treatment head also incorporates the following facilities:

• 1.

  Laser illumination: The central axis of the collimator can be illuminated using a 532 nm (cat. no. DJ532-40; Thorlabs, Newton, NJ) laser reflected off a thin 45° mirror made from an aluminized PET film (6 μm PET + ∼10 μg cm^–2 Al) in the main body of the collimator. This can be used to facilitate alignment when CBCT is not required or to help with initial positioning of the mouse prior to imaging.

• 2.

  Maxwell-type kinematic mount: Implementation enables easy removal and replacement of the collimator in a reproducible fashion and includes sensors to identify when the collimator is in position for therapeutic irradiations.

• 3.

  Bespoke collimators: The new treatment head includes a range of interchangeable collimators made from a 3.3-mm thick lead material mounted on 10 mm of brass, through which the required hole is drilled/machined. These also include an adapter, enabling the use of existing Xstrahl collimators.

Dosimetry was performed using Gafchromic™ EBT3 film (International Specialty Products, Wayne, NJ). The films were scanned as 48-bit RGB TIFF images at 300 dpi resolution using an Epson® Expression 10000 XL flatbed scanner (Plainfield, IN) at 24 h postirradiation. The dose was then calculated using optical density of the red channel and corrected using the optical density of the blue channel in conjunction with a calibration curve (16 ). Absolute dosimetry was performed following the recommendations of the report of AAPM Task Group 61 (17 ) with the EBT3 film calibrated over a range of doses (0–5 Gy) with identically filtered 220 kV X rays at 2 cm depth in Plastic Water® LR (CIRS, Norfolk, VA) using a 10 × 10 cm² field and a calibrated ionization chamber (NE2581; Thermo Scientific™, Rockford, IL). The surface dose rate along the central axis at a source-to-surface distance (SSD) of 34 cm varied from 2.4 Gy/min for a 10-mm diameter beam to 2.2 Gy/min for a 2-mm diameter beam.

The pulsatile characteristics of the breathing patterns of anesthetized mice were exploited to perform gating of the delivered beam while still obtaining a high (50–70%) efficiency of delivery. A signal representing the breathing pattern was obtained by sensing the diffuse optical reflectance using a pair of closely spaced 1-mm polymer optical fibers placed close to, but not in contact with, the skin surface of the animal. One of the fibers delivered light at 820 nm, provided by a light-emitting diode (LED) modulated (100%) by a 1 kHz square wave. The output of the other fiber was sensed with a photodiode. After trans-impedance amplification of the photodiode current, the resulting signal was passed through a synchronous rectifier, also driven by the LED modulation signal: the system was thus made insensitive to other steady-state light sources or sources modulated at frequencies other than the 1 kHz modulation. The synchronously rectified reflectance signal was split into two paths: one was heavily low-pass filtered (0.01 Hz cut-off frequency) to derive the average reflectance level; this is used as part of an automatic gain control loop to attenuate the LED output if required, should the fiber-animal distance be too low. The dynamic range is thus increased, making the system somewhat insensitive to the precise fiber pair-to-animal surface distance. The other path was low-pass filtered with a third-order Bessel filter (15 Hz–3 dB frequency) and a second-order high-pass Butterworth filter (–3 dB frequency = 0.2 Hz). The resulting pulsatile waveform was displayed in near-real time using a customized analogue-to-digital converter arrangement and a sampling rate of 100 μs. The digital data stream was used to: 1. determine the average breathing rate (average of 8 inter-breath periods), also displayed continuously; and 2. derive a gating signal.

In practice, the fiber pair can be used “head on” or can be fashioned to act as a “side-viewing” system, and placed near-parallel to the animal. The latter arrangement is useful in avoiding the collimator or when used in an animal cradle compatible with other small bore imaging devices, as shown in Fig. 6. The gating signal itself is derived by first thresholding the breathing signal and feeding the resulting pulse waveform to a microprocessor that used an adaptive algorithm (Fig. 7). This algorithm determines the duration of the breathing pulse and the inter-breath period. It also adds a “safety margin” around each breath (inhalation–exhalation period), typically adding between 50–200 ms before inhalation and after exhalation periods. The algorithm also takes into account the potential changes in the inter-breath periods (e.g., when the instantaneous breathing rate of the mouse is not constant). It delivers a gating signal that adapts to maximize the length of the minimal-movement periods of the animal surface. This is shown in Fig. 7E, where the first period shows the response when the timer E is adjusted correctly; the second period shows the situation when the next breathing cycle is late; the third period shows the situation when the next breathing cycle is early. In this last situation, some slight degradation in the precision of the radiation delivery is unavoidable. This gating signal is used to operate the fast-acting X-ray shutter described earlier.

Of note, the Bessel low-pass filter exhibits a group delay of just under 20 ms between actual movement and the displayed respiratory trace. The additional safety margins should thus, in principle, be applied more generously at the start of the respiratory cycle rather than at its end. Of course, this also depends on the specific trigger level used. Nevertheless, the animal's tissue compliance can be considered to be rather high. After an inhalation–exhalation “impulse”, a longer safety margin at the impulse end may be considered useful.

Experiments were performed to test the ability of the optical respiratory monitoring system in conjunction with the fast shutter to successfully provide respiratory-gated beams, with the delivery of X rays restricted to the period during the breathing cycle where movement is minimal. These experiments were performed using a C57BL/6 mouse anesthetized by isoflurane in oxygen-supplemented air (30% O₂), with the optical fibers used to monitor respiration positioned just above the mouse to detect chest movement. To achieve good image contrast and enable quantification of internal movement within the abdomen, a 3.2-mm diameter ball bearing (BB) was surgically implanted just below the liver. The mouse was imaged from the side at 120 frames per second using 100 kV X rays in conjunction with a fast X-ray camera [Flea3 FL3-FW-03S1M (Point Grey Research Inc./FLIR®, Richmond, Canada), used with a DRZ-std Phospor screen (MCI Optonix, Sedona, AZ), covered with a carbon fiber sheet]. Movies were collected with and without gating of the imaging X-ray beam for pre- and post-inhalation margins varying from 50–200 ms and breathing rates from ∼100 BPM down to ∼ 36 BPM, the latter achieved by varying the level of isoflurane from ∼1% to ∼2%. The movement was subsequently analyzed using the MATLAB® function “imfindcircles” (MathWorks® Inc., Natick, MA), which used a two-staged circular Hough transform (18, 19 ) to determine the displacement of the center of the BB as a function of time.

In a second experiment, a small rectangular sheet of EBT3 Gafchromic film (Ashland® Inc., Covington, KY) was supported vertically on the moving chest of the mouse using a 3D-printed support. This film was irradiated with a ∼1-mm diameter 220 kV X-ray beam with respiratory gating (50 ms pre- and post-inhalation margins) at a breathing rate of ∼100 BPM. These films were scanned at 24 h postirradiation, as described in the previous section: Dosimetry.

In vivo experiments were performed according to the guidelines of University of Oxford and within the limits of the Project License issued by the Home Office, United Kingdom for animal welfare.

Respiratory gated magnetic resonance imaging (MRI) illustrates the internal motion that occurs within a mouse over the breathing cycles (Fig. 8, Supplementary Movie S1; http://dx.doi.org/10.1667/RR14753.1.S2). The extent of the motion is dependent on the location of the target, but movement of the order of 5 mm was observed in the lung and diaphragm close to the spine. This movement was also reflected in the fluoroscopy X-ray images taken using the main SARRP PerkinElmer imaging plate, captured at a sub-Nyquist rate of 6 fps during the breathing cycle (Fig. 9A and B, Supplementary Movie S2A; http://dx.doi.org/10.1667/RR14753.1.S3). With an anesthetized mouse, the respiration (inhalation followed by exhalation) was followed by an extended rest period, where organ motion is minimal compared to the significant movement associated with breathing (as shown in Supplementary Movie S1). Due to the movement of the lung and abdomen during the breathing cycle, treatment of a region within this volume would be problematic with non-gated arrangements, with the targeted region moving in and out of a tightly collimated beam. This could be addressed by increasing the beam size to encompass the motion, however, doing so would result in exposure to an increased volume of the surrounding tissue. Alternatively, our respiratory gating approach can be used to restrict beam delivery to the resting phase of the breathing cycle after exhalation.

The optical respiration sensor was implemented to monitor the breathing cycle of an anesthetized mouse and gated X-ray beam delivery, with the breathing rate varied between ∼36 BPM and ∼100 BPM, depending on the isoflurane concentration. Figure 10 shows the response of the adaptive gating for pre- and post-margins varying from 50 ms to 200 ms. These data demonstrate the ability of this system to successfully restrict beam delivery (shutter open) to the resting phase of the breathing cycle and adapt to changes in breathing rate within a single breathing cycle. The period of beam delivery is determined by the previous breathing period. However, if the “next” breath occurs earlier than expected, this will be detected and the shutter will close in <20 ms, thereby minimizing the consequences of movement. Although the data presented is for a single C57BL/6 mouse, similar traces have been observed for a range of normal and tumor-bearing mice over a range of breathing rates. The direction of the peak and amplitude is dependent on the position of the optical fibers with respect to the chest or abdomen, but always shows a characteristic pulsatile peak corresponding to inhalation–exhalation breath, followed by a resting phase. These differences are addressed by manually adjusting the trigger level and margins prior to irradiation to achieve optimal gating. While some mice will have very regular respiration, others can be much more irregular, but this is effectively handled using adaptive respiratory-gating techniques detailed in section: Adaptive Breating Gating. In addition to data shown in Fig. 10, this has been extensively tested using a small loudspeaker driven at frequencies from 1–5 Hz and varying the mark-space ratio (data not shown).

The duty cycle for X-ray beam delivery (percentage of time the shutter is open) is dependent on the breathing rate and choice of margins. For the data shown in Fig. 10, the duty cycle for beam delivery was 82% at a breathing rate of 36 BPM and 50 ms margins. This decreased to 47% when the margins were increased to 200 ms (36 BPM) and to 61% when the breathing rate increased to 93 BPM (50 ms margins).

While fluoroscopic imaging at 6 fps of an anesthetized mouse using an ungated, continuous X-ray beam clearly showed significant tissue motion (Supplementary Movie S2A; http://dx.doi.org/10.1667/RR14753.1.S3), imaging through the respiratory-gated X-ray shutter essentially limited X-ray delivery to the resting phase of the breathing cycle, minimizing tissue movement during this period (Supplementary Movie S2B and S2C; http://dx.doi.org/10.1667/RR14753.1.S4 and http://dx.doi.org/10.1667/RR14753.1.S5, respectively). The ability to achieve respiratory gated X-ray delivery was further tested by imaging a ball bearing (BB) implanted in the mouse, the imaging being performed using the fast X-ray camera operated at 120 fps. The plots in Fig. 11 show the observed movement of the BB as a function of time when the mouse is breathing at ∼40 BPM and ∼100 BPM when continuously illuminated with an imaging X-ray field and when the X rays are gated using the optical gating system. The nongated data clearly show movement of the BB during the breathing cycle and is correlated to the movement of the diaphragm, although due to the position of the BB within the abdomen, the displacement is smaller than observed elsewhere in the body (Figs. 9C and D, and 11). However, when the X rays were gated, almost no movement of the BB was observed (the gaps observed in the traces reflect the period where no image was obtained due to the shutter preventing the transmission of X rays).

Respiratory-gated delivery of a treatment beam was subsequently tested using a ∼1-mm diameter beam targeted at an EBT film supported vertically on the abdomen of an anesthetized mouse breathing at 100 BPM. Figure 12 shows scans of the resulting EBT3 films, after exposure with and without respiratory gating, along with the resulting dose profile across the resulting spots. While the non-gated dose shows a blurring of the dose distribution, which would result in a reduction in dose to the targeted volume and a raised dose in the surrounding tissue, the gated distribution is significantly narrower and symmetrical. Nevertheless, it is noted that there are regions in the abdomen and lung where the movement can be significantly greater, up to the order of 5 mm, as shown in Figs. 8, 9 and Supplementary Movie S1 (http://dx.doi.org/10.1667/RR14753.1.S2). This would lead to a much more pronounced effect on the resulting dose distribution and therefore greater benefits for the implementation of respiratory-gated dose delivery. The overall benefit of respiratory gating is also dependent to some degree on the dimensions of the collimator used and the animal's breathing rate. While the use of isoflurane as an anesthetic results in a significant depression of the breathing rate, this is typically not the case with injectable anesthetics. However, the gating system can cope with higher breathing rates due to the quick response time of the shutter (<20 ms), and the system has been demonstrated to work at frequencies corresponding to a breathing rate of 300 BPM (see Supplementary Fig. S1; http://dx.doi.org/10.1667/RR14753.1.S1), well in excess of resting respiratory rates typically observed in unanesthetized mice.

Image-guided irradiation comes into its own when targeting and treating orthotopic tumors on internal sites, with treatment time imaging to determine the size and location of the tumor, allowing complex treatment planning for precise delivery of the required dose to the tumor while minimizing dose to the surrounding normal tissue, especially critical organs. These approaches are being used increasingly in preclinical studies with image-guided small animal irradiators. For some anatomic sites with no discernible movement during treatment (such as the brain) this method can be successfully applied with good accuracy. However, for other sites within the body, breathing related tumor motion may constitute a critical factor in limiting the accuracy of beam delivery in intrathoracic and intraabdominal tumors. Indeed, the movement of the target in and out of the treatment field inevitably leads to an increase in the heterogeneity of the dose across the tumor and to an associated reduction in the average dose along with increase in dose to the normal surrounding tissues (20 ). In human patients, the advent of respiratory gated image-guided radiotherapy has facilitated complex treatment planning to precisely deliver high doses to intrathoracic and intraabdominal tumors while minimizing the dose the surrounding normal tissue, especially critical organs (21 ). In contrast, in the preclinical radiobiology setting, the concept of respiratory gating to enable precise and less toxic irradiation of orthotopic and transgenic tumor mouse models remains largely unexplored due to the lack of availability of such systems and the technical challenges posed in working with small animals. In that context, we developed and implemented an optical breathing monitoring system, associated control software and fast X-ray shutter, which enables the successful delivery of respiratory-gated X-ray beams on the SARRP, thereby increasing accuracy of beam delivery in mouse models. The system is adaptive, and can therefore handle a wide range of breathing rates, as well as irregularities in the breathing period. Because of the quick response time of the shutter (<20 ms), the X rays can be quickly turned off when breathing is detected earlier than expected, thereby restricting radiation delivery to the period during the breathing cycle when tissue movement is minimal compared to the significant movement associated with breathing (as shown in Supplementary Movie S1; http://dx.doi.org/10.1667/RR14753.1.S2). Additionally, the reduced time interval is used to predict the time of the next breath. An alternative way to address this problem would be to intubate and ventilate the animal to externally control the breathing cycle so that breaths occurred only while X rays are turned off; however, this procedure requires a high level of skill and caution to prevent damage to the trachea (22 ). The rotary motion of the shutter not only means that the operation is silent, and therefore will not disturb the anesthetized animal, but will also minimize any potential movement or vibration of the head and collimator during operation (assuming the shutter assembly is firmly attached to a large, stable apparatus such as the X-ray tube). The overall benefit of respiratory gating will also be dependent on the size of collimator being used, by the breathing rate, and importantly, by the position of the target within the animal's body. With deviations of up to the order of 5 mm in places for a mouse (Figs. 8 and 9) and potentially greater deviations in rats, the use of small collimators may result in significant underdosing of the target with additional exposure of the surrounding tissue. The extent of this motion along with breathing rate could potentially be modified by the size and position of a tumor. As a result, movement will also be an important issue with the increasing use of “dose painting” approaches, the use of combinations of small fields to produce irregular-shaped dose distributions to a target volume or non-uniform dose distribution (23, 24 ). The effects are likely to become more significant with increasing breathing rates, due to a decrease in the fraction of time during the breathing cycle taken by the stationary resting phase. The benefit of respiratory gated beam delivery has also been illustrated by theoretical calculations, which predicted that respiratory motion could result in an 11% reduction in mean dose to a lung tumor, located near the diaphragm, compared to the planned dose assuming no motion (25 ). Therefore, even if respirator-gated delivery is not possible, it is important to assess the potential implications of any respiration-related movement on the resulting dose distributions.

Although CBCT is intended to enable accurate targeting, its inherent low soft tissue contrast does not enable identification of the abdominal organs. This is currently being addressed at Oxford by using T2-weighted MR images to enable identification and targeting of sites, which are then co-registered using nonrigid registration techniques with CBCT images subsequently obtained using the SARRP (26 ). An animal holder has been designed, which can be transferred between the MR system and the SARRP, and which also incorporates the optical fibers used for monitoring breathing, allowing respiratory-gated MR images to be obtained. This enables us to determine the position of the tumor and surrounding organs during the rest phase of the breathing cycle. Gated X-ray delivery is also restricted to this period. In connection with MRI, the breathing rate monitoring optical detection hardware has the potential to be used for detecting heart rate information as well, on a nude or partial-shaved rodent. However, heart rates can reach 20 Hz, which is not only too fast for our gating system to handle, but also one where the duty cycle would be far too low. Therefore, heart rate gated X-ray beam delivery is not possible with the current system, though it can be used for MR image acquisition.

Gated beam delivery necessarily reduces the dose rate and this should be taken into account. Significant reduction of average dose rates should be avoided to ensure that the total beam delivery time is significantly less than repair times, due to the potential decrease in biological effectiveness (27 ). This limitation ultimately dictates the smallest collimator dimensions. In our system, we can typically achieve a nongated dose rate of 2.2 Gy/min with a 2-mm diameter beam, i.e., 1.1 Gy/min with a 50% dose delivery duty cycle. It would indeed be beneficial to develop systems that rely on beam tracking rather than gating, but this is a difficult engineering challenge, particularly on current commercial platforms.

Other features of the new X-ray treatment head have been beneficial when using the SARRP irradiator for routine, nongated irradiation of small animals. The use of a Maxwell-type kinematic mount enables the easy removal (for CBCT imaging) and replacement (for irradiation) of the collimator in a reproducible fashion. The addition of the programmable filter drive helps to minimize “mistakes” when changing beam-hardening filters between imaging and irradiation phases: whenever the collimator assembly is removed, the beam-hardening filter is also moved out of the way. Additional software-selected filter positions are also available if different beam-hardening filters are required. The transmission ionization chamber, which is locally corrected for air density, not only enables a quantitative check of X-ray delivery, but can also be calibrated so that the shutter, and therefore beam delivery, can be defined and controlled using monitor units rather than irradiation time. This corrects for ageing tube output, though it is of course essential to ensure correct tube-collimator alignment to obtain the maximal dose rate. Also, the target alignment laser has been found useful to align the mouse to a collimated beam manually, when CBCT is not required, and helps with the initial positioning of the mouse prior to imaging.

While respiratory gated dose delivery is specifically discussed here, this system also has the potential for use in improving the on-board imaging by giving the option to acquire a respiratory-gated CBCT. However, the current PerkinElmer imaging panel acquisition is free-running and not synchronized to animal stage rotation making, reducing the apparent focus of each acquired projection. Furthermore, the CBCT reconstruction algorithm is inherently tolerant to such acquisition tolerance. Retrospective gating could be applied, but our experience to date has shown no significant benefit with the current setup of no stage-panel synchronization.

In summary, the development of the new treatment head has improved the reliability and ease of operation (automated filter selection and positioning laser) of the SARRP irradiator, in addition to the ability to monitor beam delivery (transmission ionization chamber). The development of a silent, fast X-ray shutter, along with the optical breath monitoring technology and associated adaptive gating control, is presented. This can be used to restrict beam delivery to the resting period of the breathing cycle, where movement is minimal. The arrangement is capable of detecting and quickly responding to changes in breathing rate. The development and implementation of these respiratory gated irradiation techniques facilitates improved dose delivery in the chest and abdomen of rodents where breathing related motion can be significant, helping to ensure optimal dose delivery to the target while minimizing dose to surrounding tissue. This in-house-developed system has proven to be a reliable and important new tool for preclinical radiation studies for orthotopic tumors or other treatments targeted at the animal trunk. A modified version of this system is now commercially available for the SARRP from Xstrahl.

Fig. S1. Shutter response for a 10% mark-space ratio, 25 ms on pulse width, acquired at 40 ms/div horizontal time base. Top trace: command pulse; bottom trace: delayed shutter response. These traces were taken with the shutter operated repetitively at 5 Hz and a command to open for 20 ms. They demonstrate that the short (<20 ms) latency times at shutter energization and de-energization allow operation at 5 Hz or even 10 Hz, breathing rates in excess of those typically associated with unanesthetized animals. However, the useful duty cycle would then be reduced, and a higher dose rate source may be required for appropriate radiobiological operating conditions.

Movie S1. Respiratory gated T2 MR movie of a C57BL/6 mouse (obtained using a Varian 7.0 T scanner) displays the motion observed over a complete breathing cycle of ∼1 s. The sequence is triggered to start just prior to the onset of the movement resulting from inhalation, and follows the movement through to the subsequent resting phase after exhalation. The mouse is lying prone, with its head to the right.

Movie S2. Fluoroscopy X-ray imaging of an anesthetized mouse captured at a sub-Nyquist rate of 6 fps with the PerkinElmer imaging panel during the breathing cycle using an uncollimated X-ray field. Film clip A: Constant imaging through an open shutter. Film clip B: Imaging through a respiratory gated shutter (the dark film segments represent the periods where the shutter is closed. Film clip C. The same movie shown in film clip B after the sections where the shutter is closed are removed, to demonstrate minimal movement during X-ray delivery. The turn-on/off times for an actual treatment beam will be significantly reduced due to the use of a tightly collimated beam. Imaging is performed in a free-running rate and the use of sub-Nyquist imaging rate inevitably results in some shutter transitions being imaged, causing apparent partial frames to be acquired. Information on movement during the inhalation-exhalation breath is limited due to the low frame rate of imaging panel.

The authors acknowledge the expert assistance of the Mechanical and Electronic Workshops of the CR-UK/MRC Oxford Institute for Radiation Oncology. Funding from Cancer Research UK (programme grant nos. C5255/A15935, C5255/A12678) and the Medical Research Council Strategic Partnership Funding (MRC-PC-12004) for the CRUK/MRC Oxford Institute for Radiation Oncology Cancer Imaging Centre (grant no. C2522/A10339) is gratefully acknowledged.