A four-dimensional controller for DMLC-based tumor tracking

Abstract

The development of four-dimensional (4D) radiotherapy for dynamic multileaf collimator (DMLC)-based tumor tracking requires the development of new hardware and software for controlling the radiation delivery, termed a four-dimensional controller (4DC). The four-dimensional controller integrates the information from the treatment plan and the respiration signal and computes multileaf collimator positions which are then sent to the DMLC to facilitate tumor tracking. The purpose of this study was to (a) develop a prototype 4DC and (b) determine the efficacy of the 4DC for 4-dimensional radiotherapy. In this work, three key components of 4DC were studied: (1) the effect of respiratory coaching on the reproducibility of respiratory patterns, (2) the mechanical ability of the DMLC to perform tumor tracking and (3) the ability to predict respiratory patterns to account for the delay between respiratory signal acquisition and DMLC. (1) Respiratory coaching: Respiration reproducibility is an important component, not just of 4D radiation delivery, but also 4DCT acquisition and 4D treatment planning. To investigate if coaching methods can improve respiration reproducibility, 300 four-minute respiratory traces were obtained from 23 lung cancer patients enrolled in an IRB-approved study under the conditions of free breathing, audio coaching, and audio-visual coaching. Respiration reproducibility was quantified by cycle-to-cycle position, displacement and breathing period variations. (2) DMLC capability: To investigate the mechanical capabilities of the DMLC to perform tumor tracking for respiratory motion, the maximum leaf velocity, acceleration and deceleration were measured for three different MLCs at different time periods and under various gravitational and friction environments. The measured values were input into equations of motion to determine the DMLC response time for positional changes under various initial and final conditions. The DMLC response was compared with fluoroscopy-measured diaphragm motion from 60 data-sets from five lung cancer patients. (3) Respiration prediction: To account for the system response time between respiration signal acquisition and execution of the MLC motion, a linear adaptive prediction filter was employed. This filter was implemented and the prediction accuracy tested on sixty fluoroscopy datasets from lung cancer patients in which the diaphragm motion was tracked for system response times of 0.2, 0.4 and 0.6 seconds. (1) Respiratory coaching: Respiratory reproducibility is significantly improved with audio-visual coaching. The variation in breathing rate and position and displacement was less for audio-visual coaching than free breathing and audio coaching. (2) DMLC capability: Significant differences were observed between inner (0.5 cm thick) and outer (1.0 cm thick) leaves. The DMLC maximum leaf velocities are 3.9 ± 0.5 cm/s and 3.3 ± 0.1 cm/s respectively and acceleration 37 ± 8 cm/s2 and 37 ± 10 cm/s2. Based on these measurements, the mechanical MLC delay time is around 0.1 seconds leading to an overall system response time of approximately 0.4 seconds. Applying equations of motion derived from these measurements to fluoroscopy-measured internal motion indicates that DMLC tracking will be over 95% efficient, however a beam-hold is required to account for irregular respiration such as coughing. (3) Respiration prediction: The ability to predict respiratory motion as the system response time increases. However, for a response time of 0.4 seconds (the estimated value for 4D radiotherapy utilizing current technology) the position prediction error is less than 2 mm (1.s.d.). A prototype version of a four-dimensional controller for DMLC-based tumor tracking has been developed. Based on patient respiratory and fluoroscopy analysis, and DMLC mechanical measurements and analysis, the existing linear accelerator technology is capable of 4D radiation delivery.