Intensity modulated radiation therapy (IMRT) is a current technique for the treatment of cancerous tumours, constantly expanding and developing. Image guided radiation therapy (IGRT) is used to verify that the patient is correctly positioned before delivering a dose fraction. Tomotherapy is an IMRT technique that consists of delivering the dose helically with a fine beam (1 to 5 cm). It is marketed by Accuray under the TomoTherapy brand. This system has a megavoltage CT detector (MVCT) to implement the IGRT.
International guidelines recommend to verify the accuracy of the treatment dose.
A dose of quality maximises the probability of controlling the tumour while minimising the risk of damage to healthy tissues, thus maximises the chances of complication free remission for the patient.
The commonest practice is to systematically measure a dose fraction with a radiochromic film or a detector array placed in a homogeneous phantom before starting to treat the patient. This quality assurance (QA) protocol verifies in a phantom the accuracy with which the treatment unit delivers the dose. However, it does not verify the accuracy with which the treatment planning system (TPS) calculated the dose in the patient's various anatomical structures.
Additionally, this protocol does not allow the clinicians to predict the difference between delivered and planned dose in the patient's organs at risk and target volumes.
In order to meet these needs, the work done in the context of this thesis focused on the verification of the dose calculated by the TPS and on the dose delivery quality assurance (DQA) in tomotherapy.
Firstly, an independent dose calculation software, CheckTomo, was upgraded following the launch of TomoEDGE. TomoEDGE allows a better longitudinal conformation of the dose, thus preserving healthy tissues located at the tumour's front and back. CheckTomo was tested on TomoPhant plans (used for routine QA measurements in a phantom) with dynamic jaws and on 30 clinically accepted plans. In the target volume of the TomoPhant plans, dose calculation errors up to 5 % were observed. The clinical plans were subjected to gamma-index pass rate tests. With tolerances of 3 %/2 mm (global normalisation), the pass rate was less than 95 % in 53 % of the cases. The gamma-index pass rates on plans with dynamic jaws were on average the same as on plans with static jaws. This suggests that CheckTomo's overall low accuracy does not depend -- or not only -- on the jaw mode. Finally, an overall dose error of 3 % was applied to the plans. In this case, all plans failed the gamma-index pass rate test with tolerances of 3 %/ 2 mm and a threshold of 95 %. We conclude that CheckTomo is highly sensitive to global errors.
Secondly, the open times of the collimator leaves (LOT) were measured and the delivered dose was calculated in the patient's planning CT images. On tomotherapy units, the MVCT detector measures the photon fluence exiting the patient during treatment. An algorithm for measuring the LOTs based on detector data has been developed. For the dose calculation, a stand-alone calculator provided by Accuray was used. The LOT measurement algorithm was tested using data from the TomoTherapy Quality Assurance (TQA) Daily QA procedure and data from 25 clinical plans. Clinical data were collected once in air and once in vivo i.e. with the patient on the treatment couch).
As the signal strength decreases sharply as the jaws narrow for a beam with a nominal width of 5.0 cm, it was not possible to measure the LOTs below a jaw aperture of 13 mm.
For larger apertures and for beam with other nominal widths (1 and 2.5 cm), the algorithm proved to be robust.
It allowed LOTs to be measured on static or dynamic jaw plans without any variation of the uncertainty in function the beam width. Similarly, the LOTs measurement uncertainty was not significantly greater in vivo than in air. To test the feasibility of calculating the dose from the measured LOTs, random LOT "errors" (following a Gaussian distribution) were generated and introduced into a TomoPhant plan. In total, six plans with average errors of -6 %, -4 %, -2 %, 2 %, 2 %, 4 %, and 6 %, respectively, were generated. The dose of each of these plans was measured with ionisation chambers placed in a phantom at the centroid of the target volumes. The LOTs were measured from the detector data and the delivered dose was calculated. Measured doses and calculated doses corresponded within 0.5 %, indicating a good reliability of the measured LOTs. Finally, on clinical cases, a correlation of 0.84 was observed between the median relative LOT error and the dose change in the target volume. The median relative LOT error could therefore be an easily measurable indicator of the dose delivery quality.
An DQA protocol based on the LOT measurement could reduce the workload related to IMRT quality assurance by eliminating the need to measure the delivered dose in a phantom. In addition, measuring in vivo the LOTs during the delivery of each dose fraction would allow the medical physicists to perform a DQA of each fraction individually. Combining the on-line LOT measurement with an independent dose calculation algorithm would allow dose calculation and delivery QA to be performed jointly. In addition, calculating the dose in the patient's daily images would make it possible to set up a DQA protocol very close to the in vivo dosimetry.