Dose-guided radiotherapy for lung tumors

Abstract

The transit in vivo dosimetry performed by an electronic portal imaging device (EPID) is a very practical method to check error sources in radiotherapy. Recently, the present authors have developed an in vivo dosimetry method based on correlation functions, F (w, L), defined as the ratio between the transit signal, S(t) (w, L), by the EPID and the mid-plane dose, D(m) (w, L), in a solid water phantom as a function of the phantom thickness, w, and of the field dimensions, L. In particular, generalized correlation functions F (w, L) for 6, 10 and 15 MV X-ray beams supplied by a pilot Varian linac, are here used by other three linacs operating in two centers. This way the workload, due to measurements in solid water phantom, needed to implement the in vivo dosimetry method was avoided. This article reports a feasibility study on the potentiality of this procedure for the adaptive radiotherapy of lung tumors treated by 3D conformal radiotherapy techniques. In particular, the dose reconstruction at the isocenter point D(iso) in the lung tumor has been used as dose-guided radiotherapy (DGRT), to detect the inter-fraction tumor anatomy variations that can require new CT scans and an adaptive plan. When a difference greater than 6% between the predicted dose by the treatment planning system (TPS), D (iso,TPS) and the D(iso) was observed, the clinical action started to detect possible anatomical lung tumor changes. Twelve over twenty patients examined presented in vivo dose discrepancies due to the tumor morphological changes during treatments, and these results were successively confirmed by new CT scans. In this work, for a patient that showed for all beams, D (iso) values over the tolerance level, the new CT scan was used for an adaptive plan. The lung dose volume histogram for D (iso,TPS) = 2 Gy per fraction suggested the adaptive plan. In particular, the lung volume included in 2 Gy increased from 350 cm(3) of the original plan to 550 cm(3) of the hybrid plan, while for the adaptive plan the lung volume included in 2 Gy decreased to 15 cm(3). Moreover, the mean doses to the organs at risk were reduced to 70%. The results of this research show that the DGRT procedure by the D(iso) reconstruction, integrated with radiological imaging, was feasible for periodic investigation on morphological lung tumor changes. This feasibility study takes into account the accuracy of two algorithms based on the pencil beam and collapsed cone convolution models for dose calculations where large density inhomogeneities are present.

A Feasibility Study for in vivo Dosimetry Procedure in Routine Clinical Practice

“… 2 – 6 The in vivo dosimetry (IVD) is strongly recommended by various organizations 7 - 10 so that in some countries, IVD tests are required as a routine check to detect the causes of dosimetric errors. 11 , 12 Several researchers 13 - 35 have implemented IVD systems to reconstruct the delivered dose to the patients using the amorphous silicon electronic portal imaging device (aSi-EPID) associated with the linear accelerator (LINAC). The results showed that aSi-EPIDs can be used to reconstruct the dose in single points, 2 dimensional (2D), and entire 3-dimensional (3-D) dose distribution.…”

A generalized calibration procedure for in vivo transit dosimetry using siemens electronic portal imaging devices

“…Finally, the present method can detect errors due to organ motion and patient morphologic variations by means of the EPID signal profiles [25,26], but it does not allow a complete dosimetric and geometric verification, because, without contextual CT data, the contributions to an observed deviation between planned and measured doses may not always be recognized. For instance, patient weight loss may cause similar dosimetric deviations as a machine error.…”

A stochastic model for tumor geometry evolution during radiation therapy in cervical cancer

“…This probability map might find application in optimization approaches where voxel-specific doses or weighting factors are used to plan and adapt treatments. [18][19][20][21][22][23][24] Deformation was not explicitly modeled, though we recognize that substantial deformation (which was not present in our dataset) would affect our models. Third, our model aims to predict a binary image that aligns with the actual tumor image.…”