A model-based patient selection tool to identify who may be at risk of exceeding dose tolerances during pancreatic SBRT.

PURPOSE: Locally advanced pancreatic cancer (LAPC) patients are prone to experience daily anatomical variations, which can lead to additional doses in organs-at-risk (OAR) during SBRT. A patient selection tool was developed to identify who may be at risk of exceeding dose tolerances, by quantifying the dosimetric impact of daily variations using an OAR motion model. MATERIALS AND METHODS: The study included 133 CT scans from 35 LAPC patients. By following a leave-one-out approach, an OAR motion model trained with the remaining 34 subjects variations was used to simulate organ deformations on the left-out patient planning CT anatomy. Dose-volume histograms obtained from planned doses sampled on simulated organs resulted in the probability of exceeding OAR dose-constraints due to anatomical variations. Simulated probabilities were clustered with a threshold per organ according to clinical observations. If the prediction of at least one OAR was above the established thresholds, the patient was classified as being at risk. RESULTS: Clinically, in 20/35 patients at least one OAR exceeded dose-constraints in the daily CTs. The model-based prediction had an accuracy of 89%, 71%, 91% in estimating the risk of exceeding dose tolerances for the duodenum, stomach and bowel, respectively. By combining the three predictions, our approach resulted in a correct patient classification for 29/35 patients (83%) when compared with clinical observations. CONCLUSIONS: Our model-based patient selection tool is able to predict who might be at risk of exceeding dose-constraints during SBRT. It is a promising tool to tailor LAPC treatments, e.g. by employing online adaptive SBRT; and hence, to minimize toxicity of patients being at risk.

Modeling daily changes in organ-at-risk anatomy in a cohort of pancreatic cancer patients.

PURPOSE: To characterize daily geometrical variations of gastrointestinal organs with respect to pancreatic tumors, through a population-based statistical model. MATERIALS AND METHODS: The study included 131 CT scans from 35 pancreatic cancer patients treated with Stereotactic Body Radiotherapy (SBRT). For each patient, day-to-day anatomical variations of the stomach, the duodenum and the bowel were assessed from the deformation vector fields (DVF) obtained by non-rigidly registering the contours of the fractions to the planning CT scans. For the whole population, day-to-day motion-deformation patterns were abstracted using principal component analysis (PCA) on the set of DVFs mapped on a reference patient. Based on these geometrical variations, anatomies were generated to create population-based dose-volume histograms (DVH) per patient, which were also compared to clinical values. RESULTS: Through PCA, the most dominant directions of daily deformations were localized in the abdominal organs. Common patterns were found, such as stomach contraction-expansion in the anterior-posterior direction ranging from 5 to 13 mm, and superior-inferior deformations on the bowel from 7 to 14 mm. The duodenum resulted to move laterally, but in a lesser extent (4-8 mm). The population-based DVHs derived from the model mostly included the daily DVHs observed in the clinic (in >90% of the cases). CONCLUSIONS: Anatomical variations influence the delivered doses to healthy organs during SBRT. A motion model was successfully built and explored to extract the larger directions of movement of the gastrointestinal organs. Day-to-day motion modeling can potentially be used to account for geometrical uncertainties in future plan optimization and in online adaptive strategies.

Quantification of intra-fraction changes during radiotherapy of cervical cancer assessed with pre- and post-fraction Cone Beam CT scans.

BACKGROUND AND PURPOSE: With the introduction of Intensity Modulated Radiotherapy (IMRT) and image-guided plan-of-the-day strategies, the treatment of cervical cancer has become more sensitive to intra-fraction uncertainties. In this study we quantified intra-fraction changes in cervix-uterus shape, bladder and rectum filling, and patient setup using pre- and post-fraction CBCT scans. MATERIALS AND METHODS: A total of 632 CBCT scans were analyzed for 16 patients with large tip-of-uterus displacement (>2.5 cm) measured in an empty and full bladder CT scan. In all scans, the bladder, cervix-uterus, and rectum were delineated. For rectum and bladder, intra-fraction volume changes were assessed. Systematic cervix-uterus intra-fraction displacements were obtained by non-rigidly aligning the pre-fraction cervix-uterus to that in the post-fraction CBCT. Intra-fraction patient setup changes were obtained by rigidly aligning pre- and post-CBCTs using the bony anatomy. RESULTS: The mean time between pre- and post-fraction CBCT scan was 20.8 min. The group-mean intra-fraction displacements averaged over the cervix-uterus were 0.1±1.4/1.8±1.5/-2.8±1.8 (LR/CC/AP) mm. The group-mean 5th and 95th percentile intra-fraction displacements were -2.3,2.1/-0.8,4.9/-5.8,0.5 (LR/CC/AP) mm. There was a significant correlation between bladder inflow rate and cervix-uterus motion (r=0.6 and p<0.01). Intra-fraction changes in patient setup were 1.3/0.4/0.6 and 1.4/1.0/1.1 mm (LR/CC/AP), for systematic and random changes, respectively. CONCLUSION: Intra-fraction cervix-uterus motion can be considerable and should be taken into account using appropriate PTV margins.

Treatment simulations with a statistical deformable motion model to evaluate margins for multiple targets in radiotherapy for high-risk prostate cancer.

BACKGROUND AND PURPOSE: Deformation and correlated target motion remain challenges for margin recipes in radiotherapy (RT). This study presents a statistical deformable motion model for multiple targets and applies it to margin evaluations for locally advanced prostate cancer i.e. RT of the prostate (CTV-p), seminal vesicles (CTV-sv) and pelvic lymph nodes (CTV-ln). MATERIAL AND METHODS: The 19 patients included in this study, all had 7-10 repeat CT-scans available that were rigidly aligned with the planning CT-scan using intra-prostatic implanted markers, followed by deformable registrations. The displacement vectors from the deformable registrations were used to create patient-specific statistical motion models. The models were applied in treatment simulations to determine probabilities for adequate target coverage, e.g. by establishing distributions of the accumulated dose to 99% of the target volumes (D99) for various CTV-PTV expansions in the planning-CTs. RESULTS: The method allowed for estimation of the expected accumulated dose and its variance of different DVH parameters for each patient. Simulations of inter-fractional motion resulted in 7, 10, and 18 patients with an average D99 >95% of the prescribed dose for CTV-p expansions of 3mm, 4mm and 5mm, respectively. For CTV-sv and CTV-ln, expansions of 3mm, 5mm and 7 mm resulted in 1, 11 and 15 vs. 8, 18 and 18 patients respectively with an average D99 >95% of the prescription. CONCLUSIONS: Treatment simulations of target motion revealed large individual differences in accumulated dose mainly for CTV-sv, demanding the largest margins whereas those required for CTV-p and CTV-ln were comparable.

Adaptive radiotherapy in locally advanced prostate cancer using a statistical deformable motion model.

UNLABELLED: Daily treatment plan selection from a plan library is a major adaptive radiotherapy strategy to account for individual internal anatomy variations. This strategy depends on the initial input images being representative for the variations observed later in the treatment course. Focusing on locally advanced prostate cancer, our aim was to evaluate if residual motion of the prostate (CTV-p) and the elective targets (CTV-sv, CTV-ln) can be prospectively accounted for with a statistical deformable model based on images acquired in the initial part of treatment. METHODS: Thirteen patients with locally advanced prostate cancer, each with 9-10 repeat CT scans, were included. Displacement vectors fields (DVF) obtained from contour-based deformable registration of delineations in the repeat- and planning CT scans were used to create patient-specific statistical motion models using principal component analysis (PCA). For each patient and CTV, four PCA-models were created: one with all 9-10 DVF as input in addition to models with only four, five or six DVFs as input. Simulations of target shapes from each PCA-model were used to calculate iso-coverage levels, which were converted to contours. The levels were analyzed for sensitivity and precision. RESULTS: A union of the simulated shapes was able to cover at least 97%, 97% and 95% of the volumes of the evaluated CTV shapes for PCA-models using six, five and four DVFs as input, respectively. There was a decrease in sensitivity with higher iso-coverage levels, with a sharper decline for greater target movements. Apart from having the steepest decline in sensitivity, CTV-sv also displayed the greatest influence on the number of geometries used in the PCA-model. CONCLUSIONS: PCA-based simulations of residual motion derived from four to six DVFs as input could account for the majority of the target shapes present during the latter part of the treatment. CTV-sv displayed the greatest range in both sensitivity and precision.