Implementation and Clinical Validation of Mechanically-Assisted and Non-Invasive Ventilation (MANIV) in Radiotherapy for Breathing-related Tumor Motion Management.

Radiotherapy of mobile tumors faces many challenges due to breathing-related geometrical uncertainties. Breathing amplitude and frequency may deeply and unexpectedly vary from cycle to cycle, during a treatment fraction (intra-fraction variation) or between fractions (inter-fraction variation) [1]. In Protontherapy (PT), these uncertainties are even worsened by the proton range variations within the traversed moving tissues and the interplay effect between the tumor and spot scanning beam motions. These effects can unpredictively and severely distort dose distribution, and still limit the current indications of PT for thoracic/upper-abdomen cancers [2, 3]. Therefore, several motion mitigation strategies have been developed:

• Margin Strategy: this approach consists in calculating safety margins that encompass motion-related uncertainties computed from a prior planning 4D-CT scan. Although simple to implement, it inevitably results in futile dose exposure to organs at risk [4].
• Gating Strategy: respiratory gating consists in delivering the beam within a time-window of the breathing cycle, at the end-expiratory or inspiratory plateau, when the tumor is in a predefined stable position. It prevents potentially harmful irradiation of healthy tissues by reducing safety margins [4]. During Deep Inspiration Breath Hold (DIBH), the patient is asked to hold apneas after deep inspirations to prolong the gating windows and the time efficiency of the gating procedure. DIBH has become a standard of care for left breast radiotherapy. Indeed, in addition of freezing the tumor motion, it moves away the heart from the breast and inflates the lungs, allowing thus to reduce the dose to these critical organs at risk [5]. However, for all tumor sites ( breast, lung, liver), current beam delivery times typically entail several successive spontaneous BH to complete treatment, hence require complex management with onboard imaging to monitor the target position [6]. Moreover, repeating spontaneous DIBH requires a good patient's compliance and comprehension, which may be a barrier for some patients, and may degrade the accuracy of the gating procedure. Various techniques have been investigated to improve the tumor position reproducibility over successive BH or to increase BH duration to facilitate dose delivery [7,8,9]. However, the patient invariably remains actor of his breathing with subsequent unpredictive tumor position variations from BH to BH. As a consequence, the accuracy could suffer from residual motion and unpredictable changes during spontaneous breath-holds.
• Tracking Strategy: this approach relies on motion prediction models derived patient's real-time breathing pattern, allowing for the synchronization of the tumor motion with the beam motion. Accuray Cyberknife® is a LINAC mounted on a robotic arm designed for real-time tumor tracking. A correlation model is built between external motion continuously tracked by LEDs placed on the patient and internal tumor position, tracked periodically by orthogonal x-rays imagers. The correlation model is updated whenever deviations occur due to changes of the breathing pattern [10]. Tracking allows thus to significantly reduce the safety margins and to adapt continuously the treatment delivery to the breathing pattern [4]. However, the long delivery time of a single fraction, from 60 to 90 minutes [11], limits its current use in clinical practice. Again, erratic and non-reproducible breathing may degrade the accuracy of tracking and will require frequent updates of the motion correlation model, at the expense of even longer treatment time and discomfort for the patient.

Until now, none of the current strategy provides an entirely satisfactory solution for motion management. The more accurate a technique is, the less efficient it is (treatment time, feasibility, ease of clinical implementation), and vice-versa. By taking control of the patient breathing, MANIV could solve this complex problem. Parkes et al. showed first that MANIV can safely impose a regular breathing pattern on conscious and unsedated patients [12], and could mitigate respiratory motion [13, 14]. Our group has further investigated these ventilation techniques on healthy volunteers [15] and patients [16] to broaden their applicability to radiotherapy of moving tumors. Two ventilation modes appear to be of particular interest for radiotherapy :

• The Slow Controlled ventilation mode (SL) is a bi-level pressure mode of the mechanical ventilator that induces reproducible and repeated DIBH without active patient participation. This ventilation mode therefore offers a way to improve the efficiency and accuracy of respiratory gating. Indeed, a good physical condition of the patient, his compliance or his understanding of the instructions would no longer be necessary prerequisites for the feasibility of the treatment. Thus, by relieving the patient of his breathing control, MANIV would overcome the limitations of spontaneous DIBH and would allow a larger number of patients to benefit from this technique. Moreover, the intra- and inter-fraction baseline shift (= mean position variation over time) are reduced with MANIV compared to voluntary DIBH [15] and should improve the accuracy of the gating procedure. MANIV will thus facilitate both onboard imaging procedure for patient positioning and the beam delivery accuracy. In the context of proton therapy, freezing the tumor motion thanks to SL mode would allow to treat thoracic and abdominal tumors by drastically reducing the motion-related geometrical uncertainties that have been prohibitive until now to ensure satisfactory robustness of the planned dose distribution.
• The Volume Controlled ventilation mode (VC) constraints both breathing rate and tidal volume measured from the patient's spontaneous breathing parameters, and imposes a completely regular breathing pattern without increasing the tumor baseline shift [15,16]. Stabilization of the respiratory pattern over time would be beneficial for tracking strategy. We can hypothesize that the regular breathing and tumor motions imposed by MANIV would reduce the number of model updates and the overall treatment duration, with a substantial gain in efficiency of the technique. To a lesser extent, the accuracy of the technique would also be improved [17].

In summary, our group has already demonstrated that MANIV was feasible and safe on small cohorts of volunteers and patients, and significantly improved regularity of breathing-related motion or BH monitored by real-time dynamic MRI [15,16]. Based on these very encouraging pre-clinical results, MANIV might thus considerably simplify and improve all motion management strategies in both photon- and proton therapies. However, further clinical investigations are still required in real treatment conditions to validate its use for clinical routine. These include the clinical implementation of the ventilator in a LINAC environment, and the quantification of the added value of MANIV for the above-mentioned mitigation techniques.