Cardiovascular diseases following radiotherapy are major secondary late effects raising questions among the scientific community, especially regarding the dose-effect relationship and confounding risk factors (chemotherapy, cholesterolemia, age at treatment, blood pressure,…). Post-radiation coronary diseases are one of the main causes of cardiac morbidity. Some approximations are made when coronary doses due to radiotherapy are estimated, especially regarding the morphology. For retrospective studies with old medical records, only radiographs are usually available with sometimes some contours made with a simulator. For recent medical records, CT scans displaying the anatomy in 3D are used for radiotherapy simulation but do not allow the coronary artery visualization due to low resolution and contrast. Currently, coronary doses are barely assessed in clinical practice, and when it is done, anatomical prior knowledge is generally used. This thesis proposes an original approach based on hybrid computational phantoms to study coronary artery doses following radiotherapy for left-side breast cancer and Hodgkin lymphoma.
During the thesis, a method inserting hybrid computational phantoms in a DICOM format into the treatment planning system has been developed and validated. It has been adapted and tested in conditions where only radiographs provide anatomical information, as with old medical records for left side breast radiotherapy.
The method has also been adapted to perform precise dose reconstructions to the coronary artery for patients treated for a mediastinal Hodgkin lymphoma and diagnosed with coronary stenosis through a coroscanner. A case-control study was carried out and the risk of coronary stenosis on a coronary artery segment was assessed to be multiplied by 1.049 at each additional gray on the median dose to the coronary artery segment.
For recent medical records, coronary doses uncertainties related to an approach by anatomical prior knowledge segmentation were estimated for a left side breast radiotherapy by simulating different realistic coronary artery topologies in a single representative thorax anatomy and calculating doses due to beam sets, with and without irradiation of the internal mammary chain. The inter-topology variability of the mean dose to the most irradiated coronary artery, the left descending coronary artery, was assessed to 35% and 19% with and without the internal mammary chain irradiation, respectively; and it was of 76% and 49%, respectively, considering the dose to the most irradiated 2% of this coronary artery volume.
Finally, an order of magnitude of the differences between measurments by radiochromic films and dose calculations by the ISOgray treatment planning system in the peripheral field area, has been estimated by for both a simple configuration (parallelepiped physical phantom, homogeneous media, open square field) and a complex configuration (anthropomorphic physical phantom, heterogeneous media, rectangular tangential beams with wedge filter). These differences were judged significant essentially around the geometrical border of the irradiation field for both configuration.