Radiation oncology is a medical discipline which developed at the end of the 19th century.
Since its beginnings new radiation techniques and diagnostics tools were continuously developed worldwide.
Especially during the last decades radiotherapy has undergone a very fast technological development period enabling precise treatments. Multi-leaf collimators (MLC), ion beams, intensity modulated radiotherapy (IMRT), stereotactic radiosurgery (SRS) are examples of these techniques.
Radio-oncology strategies have also been changing during the last decades, therefore the clinical workflow should change accordingly. More precise diagnostic images are needed to support the introduction of new treatment concepts. Adaptive radiation therapy (ART) is a novel approach for treatment delivery for which it is necessary to take patient images during the course of the treatment to verify the validity of the calculated dose prior to the start of the therapy. When anatomic variations occur it might be necessary to replan and re-calculate/correct the delivered dose. The expression "adaptive treatment" indicates a timewise patient-specific treatment, where all anatomical changes should be taken into account to compute the actual dose distribution.
In an ART scenario deformable image registration (DIR) software has an important role for evaluating anatomical changes in the patient anatomy during the course of the treatment.
The image deformation field, obtained by comparing and registering the treatment planning computed tomography (CT) image with a repeated fractional scan of the patient, can be used to evaluate the real dose distribution and therefore correct it whenever necessary.
Some difficulties in ART may arise from both technical and medical points of view. From the medical point of view one of the most important issue is the organs contouring. The contour propagation by means of a deformation field makes the choice of the DIR software crucial.
In the following chapters it is described as an in-house developed DIR algorithm has been validated and benchmarked using both phantom and patient dataset. The in-house DIR algorithm showed comparable to better results when benchmarked against commercially available algorithms, i.e. 3D Slicer and iPlan respectively, on phantom datasets. The performance of DIR on patients dataset was varying depending on the tumour site. No significant improvement was shown on pelvic patients images, i.e. prostate and gynaecological cancers after DIR was performed, while a clear indication for automated contour propagation was shown on lung patient image datasets thanks to the higher image contrast.