• fibrosis
• IMRT
• lung
• NTCP
• SBRT

Lung cancer is one of the most common type of cancer around the world with over 1.5 million deaths every year. In particular, non-small cell lung cancer (NSCLC) is the most frequent and accounts for about 85% of cases in European countries.
In the treatment of lung tumors, radiation therapy (RT) plays an important role in treating NSCLC, and several studies have reported promising results, especially for patients with early stage NSCLC, who are unfit for surgery or who are medically operable but refused surgery.
The most recent technological advances in RT such as stereotactic body radiotherapy (SBRT), image-guided radiotherapy (IGRT) and highly conformal techniques such as intensity-modulated radiotherapy (IMRT), allow to irradiate small volumes to high doses with minimal exposure of the surrounding healthy tissues, increase the loco-regional control and reduce the risk and severity of late side effects of RT. These factors have a significant impact on the overall survival and progression-free survival.
The irradiation of the lung, however, can lead to complications such as radiological radiation-induced pulmonary injury (RRLI), which consists in the appearance of an increase in lung tissue density and opacity, seen on the follow-up chest computed tomography (CT).
The appearing of RRLI after RT of lung cancer is very common. It occurs typically in 62% of patients within six months after treatment (acute effects) and 91% thereafter (late effects). However, only a small number of patients develop clinical symptoms. Although RRLI is mostly asymptomatic, it can also lead to a progressive and irreversible decline in the pulmonary function. The prediction of occurrence of side effects after RT is crucial for optimization and evaluation of RT treatment plans and the long-term quality of patients life.
Normal tissue complication probability (NTCP) models describe the probability that a given dose distribution of a treatment will result in some quantifiable and unfavourable side effects to a tissue or organ.
For patients who were treated for lung cancer using RT, it is important to do an early and accurate evaluation of local recurrence. The diagnosis of lung cancer recurrence is performed by physicians from the analysis of CT scans acquired during the follow-up. The RRLI may appear as mass-like consolidation on CT so, it can be difficult to distinguish from local recurrence. Misclassification of a recurrence as RRLI can result in a missed-opportunity for curative-intent treatment. On the other hand, it may also lead to unnecessary interventions such as biopsy, chemotherapy, and even surgery; exposing patients to unnecessary risks and morbidity.
The knowledge of the regions of the lung, which are at high risk of RRLI, can support the clinicians when performing the diagnosis of recurrence from follow-up CTs.
RRLI is highly dependent on locally absorbed dose because it is originated from the replacement of the irradiated lung tissues with exudates or fibrotic tissues, and it is mostly confined into the radiation fields. Therefore, the dose response of a voxel of the patient image can be assumed to be independent from the dose to the surrounding lung volume. RRLI, then, can be described by a voxel-by-voxel NTCP model.
The thesis aims to develop a voxel-by-voxel NTCP model, which considers the spatial characteristics of the 3D dose distribution within an organ to predict occurrence and localization in space of RRLI from the dosimetric and clinical data from thirty-three patients treated with RT for lung cancer. In the model, the occurrence of RRLI in the voxel has been defined as relative increase of electronic density compared to a properly defined threshold. For evaluation of voxel change in density, the follow-up CT scans co-registered with the baseline scans using B-spline deformable registration, are used. Changes in electronic density (relative to water), between the two CT scans (pre-treatment and register CT) are calculated for every voxel inside the ipsilateral healthy lung (ipsilateral lung subtracted the gross tumour volume).
This map of NTCP of RRLI, can be visualized on the follow-up CT scan and support Radiation Oncologists in differentiating recurrent/persistent disease and RRLI during the follow-up of the patients. In addition, the model can be used to predict the probability of occurrence of fibrosis in starting from pre-treatment CT, and it allows the assessment of possible consequences from the specific RT plan for the single patient also before the treatment.