Capacitive monitoring system for real‐time respiratory motion monitoring during radiation therapy

Summary This work introduces a novel capacitive‐sensing technology capable of detecting respiratory motion with high temporal frequency (200 Hz). The system does not require contact with the patient and has the capacity to sense motion through clothing or plastic immobilization devices. Abstract Purpose This work presents and evaluates a novel capacitive monitoring system (CMS) technology for continuous detection of respiratory motion during radiation therapy. This modular system provides real‐time motion monitoring without any contact with the patient, ionizing radiation, or surrogates such as reflective markers on the skin. Materials and methods The novel prototype features an array of capacitive detectors that are sensitive to the position of the body and capable of high temporal frequency readout. Performance of this system was investigated in comparison to the RPM infrared (IR) monitoring system (Varian Medical Systems). The prototype included three (5 cm × 10 cm) capacitive copper sensors in one plane, located at a distance of 8–10 cm from the volunteer. Capacitive measurements were acquired for central and lateral‐to‐central locations during chest free‐breathing and abdominal breathing. The RPM IR data were acquired with the reflector block at corresponding positions simultaneously. The system was also tested during deep inspiration and expiration breath‐hold maneuvers. Results Capacitive monitoring system data demonstrate close agreement with the RPM status quo at all locations examined. Cross‐correlation analysis on RPM and CMS data showed an average absolute lag of 0.07 s (range: 0.03–0.23 s) for DIBH and DEBH data and 0.15 s (range: 0–0.43 s) for free‐breathing. Amplitude difference between the normalized CMS and RPM signal during chest and abdominal breathing was within 0.15 for 94.3% of the data points after synchronization. CMS performance was not affected when the subject was clothed. Conclusion This novel technology permits sensing of both free‐breathing and breath‐hold respiratory motion. It provides data comparable to the RPM system but without the need for an IR tracking camera in the treatment room or use of reflective markers on the patient.


| INTRODUCTION
External beam radiation therapy (RT) involves the precise delivery of ionizing radiation to predefined locations within the body to kill cancer cells while sparing the surrounding healthy tissue. For many sites, special attention to motion management is required to ensure accurate delivery. One of the most prevalent sources of motion is respiration, and has a prominent effect when treating breast, lung, or abdominal indications. Management of respiratory motion can result in improved targeting accuracy and reduced normal tissue toxicity, and also ameliorates imaging motion artifacts enabling improved tumor visualization and alignment. 1,2 Common methods of motion management include reduction of motion through abdominal compression, gating to a specific breathing amplitude or phase, and performing defined breath-hold maneuvers such as deep inspiration breath hold (DIBH) and deep expiration breath hold (DEBH). The DIBH method works by delivering the treatment while the patient holds their breath at the end of a deep inhalation. This method, when used during breast radiation treatment, for example, largely eliminates the breathing motion and pushes the heart further away from the radiation field for left sided breast cancer treatments. 1,3,4 For thoracic or abdominal indications, DIBH/DEBH acts to stabilize tumor motion, allowing for decreased planning and treatment margins. 1,3,5 Breath-hold techniques have shown dosimetric advantages and have become widely used. 6 In contrast, respiratory gating methods do not eliminate the breathing motion. Rather, they introduce a gating window wherein the radiation beam is delivered during a predefined phase of the breathing cycle. The aforementioned techniques require a continuous monitoring system to ensure the reproducibility of the breathing pattern or breath-hold position, and the systems employed have traditionally included: laser or optical surface scanning, spirometry, infrared marker tracking, or implanted radiofrequency transponders. 1,2,[6][7][8][9][10][11] Implanted radiofrequency transponders are used for motion management but involve surgical intervention with a chance of major or minor complications, transponder migration, and introduce imaging artifacts. 11 Spirometric methods work by voluntarily or involuntarily blocking the patient's breathing. While this may minimize motion, the approach may be limited by a patient's limited respiratory capacity, as well as the equipment costs and patient preparation time. 6 Infrared tracking devices rely on a limited number of markers placed on the patient's abdomen or thorax. Markers may be obscured from the IR camera view by patient body habitus, and often need to be placed prior to having complete knowledge of the patient's breathing habits. 2,10 While laser-or optical-based surface imaging provides a noncontact three-dimensional view of the chestwall anatomy, camera placement must allow a nonobstructed view of the patient's chestwall surface. Accurate surface imaging can be hindered by the position of the gantry/imaging arms and immobilization devices, requires the patient to be fully uncovered throughout the treatment, and may be affected by body hair. 9 Maintaining a constant and unobstructed view with either reflective marker or surface imaging methods may become more challenging with emerging noncoplanar treatments. 12,13 In this work, we present the first report of a capacitive-sensing technology capable of detecting respiratory motion. The technology extends the application of capacitive sensors described previously 14 in a geometry suitable for sensing motion in various regions of the thoracic or abdominal anatomy. In the development of the prototype device described herein, we aimed to satisfy the requirements of (a) not requiring direct contact with the patient, (b) modularity, that is, a portable device that is moveable and indexable between treatment systems and imaging couch tops, (c) capacity to sense motion through clothing or plastic immobilization devices, 14

2.A | Capacitive sensing and prototype design
The respiratory monitoring system works by tracking the position of the area of interest, for example, chest wall or abdomen. It can detect the motion of the region of interest (ROI) and provide information used for gating, or determining breath-hold amplitude. The system is comprised of thin (0.0254 mm) copper conductive sensors mounted on an acrylic horizontal plate above the patient's chest or abdominal area. The human body is naturally electrically conductive, [15][16][17] therefore placing the copper sensors close to the body forms a capacitor. In its simplest form, the capacitance of a parallel plate system follows Eq.
1. This equation shows that capacitance depends on the distance between the plates (d), the area of the capacitor plates (A) and the material between the plates which is introduced as permittivity (ϵ).
Therefore, capacitance will change as the distance between the sensor and patient changes due to breathing.  substrate for the sensors is acrylic in this prototype, in a clinical version of the device, we anticipate that this would be replaced by a rigid, minimally attenuating material such as thin carbon fiber. Additionally, the final design will be optimized to reduce the amount of carbon fiber in the beam, accommodate patient habitus, and reduce gantry clearance issues similar to existing devices such as abdominal compression devices. Further optimization would also provide flexibility by allowing for an arms-down setup. The prototype measures 60 cm by 20 cm by 41 cm in width, depth, and height, respectively.
To ensure clearance with the gantry, the prototype was placed around an anthropomorphic phantom as shown in Fig. 2. The phantom was at 95 cm SSD at xiphoid position and gantry clearance was found to be sufficient during 180 degree rotation. Additionally, CBCT images where acquired to ensure no artifacts are introduced in the presence of the copper sensors (Fig. 2). The RPM block was setup in contact with the anatomy of interest, and the CMS sensor array was located anterior to the same region. The volunteer was asked to take a deep breath that was used to synchronize both systems in postprocessing. The CMS system was setup to acquire data for 150 s at 200 Hz using in-house software. Once the CMS data acquisition was concluded, the RPM system was turned off. The CMS data were processed using an exponential weighting method 18

2.B.3 | Motion detection with obstructed view
Performance of the CMS system was tested with no direct view of the chest. The volunteer was clothed, and the CMS sensor was placed above the xiphoid, again at an 8-10 cm distance from the skin, and data were gathered for three DIBH instances. The RPM system was inoperable in this scenario due to the fact that the reflective block could not be stably positioned on clothing. Signal to noise ratio was calculated for the raw data using Eq. (2) below where A denotes the amplitude of signal or noise as specified by the subscript.
The amplitude of the signal was defined as the average change in acquired signal from exhale to inhale point (breathing amplitude) over the 120 s of data acquisition. The noise was estimated on a 0.005 s rolling window by calculating the signal change between two adjacent data points, averaged over the time series of acquired data (120 s) and rounded. This provides an understanding of the ratio between the range of signal and the amplitude of noise which can be helpful for clinical comparison and decision making. Figure 2 illustrates the gantry clearance with the prototype in place in the linac environment. A snapshot of the acquired CBCT is shown in Fig. 2(b) to show that presence of copper did not introduce image artifacts. central sensor experiments, the value increased to 98.2% and 99.8%

3.B | Free-breathing
for chest and abdominal breathing, respectively.

3.C | Deep inspiration/expiration breath-hold
Breathing traces for DIBH and DEBH are shown in Fig. 5  3.E | Effect of sensor-body separation on signal to noise ratio The modular nature of this device presents several advantages relative to IR or surface imaging camera-based systems. The device may be coupled reproducibly to a couch top and therefore may be relocated easily between treatment units, CT, PET/CT or angiography platforms. It may also be used outside of these rooms, for example, on a mock-up of a treatment couch, thereby providing an offline platform for patient education and coaching without tying up expensive capital equipment resources. The device could be produced at comparatively low cost, which is amenable to equipping multiple treatment and imaging rooms. This option is increasingly important, as techniques such as DIBH for treatment of left-sided breast cancer become more common. 19 Additionally, while the RPM system detects motion from a single plane, the CMS sensor detects one single capacitance value which is related to the average distance between the sensor and the body over the area of the sensor. As the breathing occurs, the distance between the sensor and body surface varies. The distance averaging resulting from a strip sensor geometry (rather than a point sensor)

3.D | Motion detection with obstructed view
helps detect the global respiratory motion despite the curvature of the individual's body.
The system charge/discharge time is 2 µs and draws a low current of 24 µA with the charge/discharge process, which occurs 200 times per second. Considering these variables and the fact that the dielectric material used in this case is air, no significant capacitive leakage or stability issues are observed or expected.
The capacitive system signal relies on the electrical conductivity of the body. This precludes conducting a phantom study, specifically in the chest/abdominal area. In this study, the CMS detection system was used on two volunteers to acquire proof of concept data and to investigate the viability of the design. However, our next step would  A comparison of the raw and processed signals shows a slight temporal shift in the processed data. The phase shift was measured at the maxima (end of exhale point) for the cases presented in Fig. 6 and was found to be 6.5°, 8°, and 10°for 3.5, 5.5, and 7.5 cm sensor-body separation, respectively. As a result, reducing the sensor-body distance is advised, for example to 4 cm, in clinical conditions. Considering a 40% gating window as an example 20 (144°), the introduced phase shift at low sensor-body separation would lead to a small (4.5%) uncertainty. The current prototype provides one dimensional positional information regarding respiratory motion. In concept, the system can be extended to include a larger array of sensors in different axial planes to provide a three-dimensional mapping of respiratory motion during treatment. The prototype F I G . 5. Simultaneous RPM and CMS breathing signal gathered during DIBH (a) and DEBH (b). Sensors were placed over the xiphoid process on bare chest. (c) CMS breathing signal gathered using the central sensor during DIBH with obstructed view of the chest (the volunteer was clothed).
described herein included an acrylic frame for ease of machining and construction. A clinical version will be made of carbon fiber, which is the status quo for clinical accessories in radiation therapy due to its radiotransmissive properties. Additionally, the current design only requires an area of 10 cm by 5 cm for the conductive sensor. This allows for reducing the size of the sensor platform ( Figure 1) and optimizing the design to reduce the amount of carbon fiber in the beam, accommodate larger patient habitus, and increase gantry clearance similar to existing devices such as abdominal compression devices.

| CONCLUSION
This work presents a novel noncontact and modular technology for real-time monitoring of respiratory motion. The current prototype can detect respiratory motion at different regions, providing positional data at 200 Hz readout frequency. The system is minimally intrusive as it does not require unobstructed view of the chest and can provide motion detection for extracranial lesions through fabric, or thermoplastic immobilization material. 14 Furthermore, the system requires no contact with the patient and is not anchored to a treatment room. This study acts as proof of concept and our next step would be a clinical study with a cohort of volunteers in different clinical setup positions.

ACKNOWLEDG MENTS
The authors acknowledge financial support from the Atlantic Canada Opportunities Agency (ACOA), Atlantic Innovation Fund (AIF), and Brainlab AG.

Ms. Sadeghi reports grants from Atlantic Canada Opportunities
Agency-ACOA, non-financial support, and other from Brainlab AG, during the conduct of the study. In addition, Ms. Sadeghi has a provisional patent application pending and a licensing agreement with F I G . 6. Normalized raw and postprocessed CMS signal acquired during abdominal respiration at different sensorbody distances. Signal to noise ratio for the raw signal is shown on each graph. The data are acquired at 200 Hz and processed using an exponential weighting method with a forgetting factor of 0.99 followed by a moving average filter of 10 samples (0.05 s) to reduce random noise.