| Sažetak (engleski) | Radiotherapy is a technique that utilizes ionizing radiation for cancer treatment. In this thesis, a method is presented that allows radiation fields with fluence resolution considerably higher than with standard methods. The primary limitation to fluence resolution is the width of leafs of multi-leaf collimators (MLC). With the realization that ever-narrower leafs do not always lead to the enhancement of key dosimetric indicators, manufacturers of linear accelerators seek other solutions to this problem, such as dual-layer MLC which are two MLCs mounted perpendicular to each other. This demonstrates that such technologies are viable development path of photon radiotherapy. Nonetheless, medical linear accelerators with single MLC are broadly in use, and a method to increase fluence resolution in the direction orthogonal to the leaf motion using existing technology is of general interest. A novel method of defining and delivering high-resolution fluence radiation beams is presented in this dissertation. This method, named cross-motion leaf calculator (XMLC), allows linear accelerators with single MLC to deliver clinically relevant high-resolution fluences comparable to those of dual-layer MLC linear accelerators without an expensive hardware upgrade. It integrates radiation planning, fluence processing and decomposition, MLC sequencing, and electronic portal imaging device (EPID) verification in a single software library. Key methods of the research are the use of Monte Carlo simulations (MCS) of beams and patients, nonlinear optimization for ideal fluence definition/decomposition, and beam dosimetry. Extensive Monte Carlo simulation of the interaction of radiation beam in patient’s body and CT image of the patient is used to allow the so-called ideal fluence definition. Such ideal fluences have high resolution but are not deliverable with standard modulation techniques, such as a single sliding window (SSW). Therefore, ideal fluences are decomposed into two orthogonal components that are deliverable by two fields with mutually orthogonal collimator settings with a standard single-layer MLC. The decomposition algorithm is specifically tuned so that resulting components don’t require more monitor units (MU) that standard methods, but deliver high-resolution fluence when two beams are superimposed. A single sliding window algorithm is re-implemented in an open-source library to allow transparent MLC delivery and direct programming of the MLC controller. Detailed Monte Carlo simulation of linear accelerator head is employed so that the key parameter for leaf sequencing may be determined. These key parameters are also measured directly on a linear accelerator Varian Clinac DMX. Finally, EPID was used to verify the delivery of high-resolution fluences. For this purpose, two techniques of beam modulation, XMLC and SSW, were compared against source ideal fluence prediction. The comparison was done with gamma analysis at two levels, clinical (3%/2mm) and detailed (1%/1mm). Standard dosimetric indicators were determined and compared. Dosimetric parameters used for leaf sequencing were determined with MCS and direct measurements and found to be consistent. Fluence decomposition processing was finely tuned to sensible and efficient output. Most important physical effects, such as spectrum hardening and leaf transmission, were considered and taken into account in the decomposition process. Gamma analysis of two modulation techniques, XMLC and SSW, was conducted against ideal fluence at two different detail levels. With parameters 3 %=2mm=10%, conformance percentage is 98,5±0,6 (XMLC) versus 74±3 (SSW) and with parameters 1 %=1mm=10%, conformance percentage is 87±4 (XMLC) versus 48±7 (SSW). This demonstrates considerable enhancement of fluence resolution with XMLC compared against SSW. Key dosimetric indicator, calculated from twenty anonymized CT images of cancer patients, were presented and compared between two modulation techniques (XMLC and SSW). For the same dose load of healthy organs, the following indicators show statistically significant (p=0,05) enhancement with XMLC with comparison to SSW respectively (V_1.9Gy: 98.4±0.7 vs. 94±2; V_1.95Gy: 95±2 vs. 87±5; HI: 0,11±0,01 vs 0,07±0,01). This translates to better tumor control with same normal tissue complication probability. The use of MCS is used for radiotherapy planning for a long time and is reaffirmed as an excellent choice for detailed dose calculation. Results also show that open-source methods can be successfully used as an inverse planning tool. Dosimetric parameters, determined from Monte Carlo simulation of the head of linear accelerator and direct measurements were compared and found to be consistent. Gamma comparison of two modulation techniques, XMLC and SSW, is presented. A considerable advantage of XMLC over SSW is noted, at both detail level comparison. It is also analyzed how and where gamma analysis fails and it is shown that the main source of discrepancy is inter-leaf leakage (ILL) which was not calculated in either of the compared techniques. Unlike SSW, there is a clear way to incorporate ILL in XMLC so that small fluence perturbation are inserted into one decomposition components to compensate for the ILL of the other decomposition component. Therefore, a clear improvement of the presented technique is identified. Finally, key dosimetric indicators are calculated and presented. It is shown that there is a statistically significant improvement of dosimetric indicators when XMLC was used while maintaining the level of protection of healthy organs at risk. This key result shows that, while comparable to SSW in every other aspect, XMLC offers improvement of dosimetric indicators. This potentially implies clinical benefits for treatment outcome, but this can only be proven in an independent clinical study that is outside of the scope of this research. A string of robust measuring techniques and algorithms are presented that allow the delivery of high-resolution fluences on a single MLC linear accelerator. Key processes are extensive use of Monte Carlo simulations for dose calculation and algorithm calibration, orthogonal decomposition of such fluence into two well-defined, achievable fluences that are delivered with two fields with mutually orthogonal collimator settings, using sliding window technique and leaf sequencing that allows MU efficient delivery. The presented work demonstrates that such a technique is superior with respect to classic, single field sliding window by means of measuring the output on an EPID device. It is also shown that critical setbacks, such as excessive radiation leakage and leaf transmission, which are usually associated with hyper-modulated plans can be successfully mitigated. Presented work may be used to research the dosimetric implications of the introduction of dual-layer linear accelerators into clinical practice. Potential improvements to the technique are identified and are subject to further research. |
