Comparison of the suitability of CBCT- and MR-based synthetic CTs for daily adaptive proton therapy in head and neck patients.

Cone-beam computed tomography (CBCT)- and magnetic resonance (MR)-images allow a daily observation of patient anatomy but are not directly suited for accurate proton dose calculations. This can be overcome by creating synthetic CTs (sCT) using deep convolutional neural networks. In this study, we compared sCTs based on CBCTs and MRs for head and neck (H&N) cancer patients in terms of image quality and proton dose calculation accuracy. A dataset of 27 H&N-patients, treated with proton therapy (PT), containing planning CTs (pCTs), repeat CTs, CBCTs and MRs were used to train two neural networks to convert either CBCTs or MRs into sCTs. Image quality was quantified by calculating mean absolute error (MAE), mean error (ME) and Dice similarity coefficient (DSC) for bones. The dose evaluation consisted of a systematic non-clinical analysis and a clinical recalculation of actually used proton treatment plans. Gamma analysis was performed for non-clinical and clinical treatment plans. For clinical treatment plans also dose to targets and organs at risk (OARs) and normal tissue complication probabilities (NTCP) were compared. CBCT-based sCTs resulted in higher image quality with an average MAE of 40 ± 4 HU and a DSC of 0.95, while for MR-based sCTs a MAE of 65 ± 4 HU and a DSC of 0.89 was observed. Also in clinical proton dose calculations, sCTCBCT achieved higher average gamma pass ratios (2%/2 mm criteria) than sCTMR (96.1% vs. 93.3%). Dose-volume histograms for selected OARs and NTCP-values showed a very small difference between sCTCBCT and sCTMR and a high agreement with the reference pCT. CBCT- and MR-based sCTs have the potential to enable accurate proton dose calculations valuable for daily adaptive PT. Significant image quality differences were observed but did not affect proton dose calculation accuracy in a similar manner. Especially the recalculation of clinical treatment plans showed high agreement with the pCT for both sCTCBCT and sCTMR.

Comparison of CBCT based synthetic CT methods suitable for proton dose calculations in adaptive proton therapy.

In-room imaging is a prerequisite for adaptive proton therapy. The use of onboard cone-beam computed tomography (CBCT) imaging, which is routinely acquired for patient position verification, can enable daily dose reconstructions and plan adaptation decisions. Image quality deficiencies though, hamper dose calculation accuracy and make corrections of CBCTs a necessity. This study compared three methods to correct CBCTs and create synthetic CTs that are suitable for proton dose calculations. CBCTs, planning CTs and repeated CTs (rCT) from 33 H&N cancer patients were used to compare a deep convolutional neural network (DCNN), deformable image registration (DIR) and an analytical image-based correction method (AIC) for synthetic CT (sCT) generation. Image quality of sCTs was evaluated by comparison with a same-day rCT, using mean absolute error (MAE), mean error (ME), Dice similarity coefficient (DSC), structural non-uniformity (SNU) and signal/contrast-to-noise ratios (SNR/CNR) as metrics. Dosimetric accuracy was investigated in an intracranial setting by performing gamma analysis and calculating range shifts. Neural network-based sCTs resulted in the lowest MAE and ME (37/2 HU) and the highest DSC (0.96). While DIR and AIC generated images with a MAE of 44/77 HU, a ME of -8/1 HU and a DSC of 0.94/0.90. Gamma and range shift analysis showed almost no dosimetric difference between DCNN and DIR based sCTs. The lower image quality of AIC based sCTs affected dosimetric accuracy and resulted in lower pass ratios and higher range shifts. Patient-specific differences highlighted the advantages and disadvantages of each method. For the set of patients, the DCNN created synthetic CTs with the highest image quality. Accurate proton dose calculations were achieved by both DCNN and DIR based sCTs. The AIC method resulted in lower image quality and dose calculation accuracy was reduced compared to the other methods.

Multi-organ segmentation of the head and neck area: an efficient hierarchical neural networks approach.

PURPOSE: In radiation therapy, a key step for a successful cancer treatment is image-based treatment planning. One objective of the planning phase is the fast and accurate segmentation of organs at risk and target structures from medical images. However, manual delineation of organs, which is still the gold standard in many clinical environments, is time-consuming and prone to inter-observer variations. Consequently, many automated segmentation methods have been developed. METHODS: In this work, we train two hierarchical 3D neural networks to segment multiple organs at risk in the head and neck area. First, we train a coarse network on size-reduced medical images to locate the organs of interest. Second, a subsequent fine network on full-resolution images is trained for a final accurate segmentation. The proposed method is purely deep learning based; accordingly, no pre-registration or post-processing is required. RESULTS: The approach has been applied on a publicly available computed tomography dataset, created for the MICCAI 2015 Auto-Segmentation challenge. In an extensive evaluation process, the best configurations for the trained networks have been determined. Compared to the existing methods, the presented approach shows state-of-the-art performance for the segmentation of seven different structures in the head and neck area. CONCLUSION: We conclude that 3D neural networks outperform the most existing model- and atlas-based methods for the segmentation of organs at risk in the head and neck area. The ease of use, high accuracy and the test time efficiency of the method make it promising for image-based treatment planning in clinical practice.

Characterizing the modulation transfer function (MTF) of proton/carbon radiography using Monte Carlo simulations.

PURPOSE: To characterize the modulation transfer function (MTF) of proton/carbon radiography using Monte Carlo simulations. To assess the spatial resolution of proton/carbon radiographic imaging. METHODS: A phantom was specifically modeled with inserts composed of two materials with three different densities of bone and lung. The basic geometry of the phantom consists of cube-shaped inserts placed in water. The thickness of the water, the thickness of the cubes, the depth of the cubes in the water, and the particle beam energy have all been varied and studied. There were two phantom thicknesses considered 20 and 28 cm. This represents an average patient thickness and a thicker sized patient. Radiographs were produced for proton beams at 230 and 330 MeV and for a carbon ion beam at 400 MeV per nucleon. The contrast-to-noise ratio (CNR) was evaluated at the interface of two materials on the radiographs, i.e., lung-water and bone-water. The variation in CNR at interface between lung-water and bone-water were study, where a sigmoidal fit was performed between the lower and the higher CNR values. The full width half-maximum (FWHM) value was then obtained from the sigmoidal fit. Ultimately, spatial resolution was defined by the 10% point of the modulation-transfer-function (MTF10%), in units of line-pairs per mm (lp/mm). RESULTS: For the 20 cm thick phantom, the FWHM values varied between 0.5 and 0.7 mm at the lung-water and bone-water interfaces, for the proton beam energies of 230 and 330 MeV and the 400 MeV/n carbon beam. For the 28 cm thick phantom, the FWHM values varied between 0.5 and 1.2 mm at the lung-water and bone-water interface for the same inserts and beam energies. For the 20 cm phantom the MTF10% for lung-water interface is 2.3, 2.4, and 2.8 lp/mm, respectively, for 230, 330, and 400 MeV/n beams. For the same 20 cm thick phantom but for the bone-water interface the MTF10% yielded 1.9, 2.3, and 2.7 lp/mm, respectively, for 230, 330, and 400 MeV/n beams. In the case of the thicker 28 cm phantom, the authors observed that at the lung-water interface the MTF10% is 1.6, 1.9, and 2.6 lp/mm, respectively, for 230, 330, and 400 MeV/n beams. While for the bone-water interface the MTF10% was 1.4, 1.9, and 2.9 lp/mm, respectively, for 230, 330, and 400 MeV/n beams. CONCLUSIONS: Carbon radiography (400 MeV/n) yielded best spatial resolution, with MTF10% = 2.7 and 2.8 lp/mm, respectively, at the lung-water and bone-water interfaces. The spatial resolution of the 330 MeV proton beam was better than the 230 MeV proton, because higher incident proton energy suffer smaller deflections within the patient and thus yields better proton radiographic images. The authors also observed that submillimeter resolution can be obtained with both proton and carbon beams.

Dosimetric assessment of a novel metal artifact reduction method in CT images.

The aim of this study was to assess the ability of metal artifact reduction (MAR) algorithm in restoring the CT image quality while correcting the tissue density information for the accurate estimation of the absorbed dose. A phantom filled with titanium (low-Z metal) and Cerrobend (high-Z metal) inserts was used for this purpose. The MAR algorithm was applied to phantom's CT dataset. Static intensity-modulated radiation therapy (IMRT) plans, including five beam angles, were designed and optimized on the uncorrected images to deliver 10 Gy on the simulated target. Monte Carlo dose calculation was computed on uncorrected, corrected, and ground truth image datasets. It was firstly verified that MAR methodology was able to correct HU errors due to the metal presence. In the worst situation (high-Z phantom), the image difference, uncorrected ground truth and corrected ground truth, went from -4.4 ± 118.8 HU to 0.4 ± 10.8 HU, respectively. Secondly, it was observed that the impact of dose errors estimation depends on the atomic number of the metal: low-Z inserts do not produce significant dose inaccuracies, while high-Z implants substantially influence the computation of the absorbed dose. In this latter case, dose errors in the PTV region were up to 23.56% (9.72% mean value) when comparing the uncorrected vs. the ground truth dataset. After MAR correction, errors dropped to 0.11% (0.10% mean value). In conclusion, it was assessed that the new MAR algorithm is able to restore image quality without distorting mass density information, thus producing a more accurate dose estimation.

Comparison between infrared optical and stereoscopic X-ray technologies for patient setup in image guided stereotactic radiotherapy.

PURPOSE: To compare infrared (IR) optical vs. stereoscopic X-ray technologies for patient setup in image-guided stereotactic radiotherapy. METHODS AND MATERIALS: Retrospective data analysis of 233 fractions in 127 patients treated with hypofractionated stereotactic radiotherapy was performed. Patient setup at the linear accelerator was carried out by means of combined IR optical localization and stereoscopic X-ray image fusion in 6 degrees of freedom (6D). Data were analyzed to evaluate the geometric and dosimetric discrepancy between the two patient setup strategies. RESULTS: Differences between IR optical localization and 6D X-ray image fusion parameters were on average within the expected localization accuracy, as limited by CT image resolution (3 mm). A disagreement between the two systems below 1 mm in all directions was measured in patients treated for cranial tumors. In extracranial sites, larger discrepancies and higher variability were observed as a function of the initial patient alignment. The compensation of IR-detected rotational errors resulted in a significantly improved agreement with 6D X-ray image fusion. On the basis of the bony anatomy registrations, the measured differences were found not to be sensitive to patient breathing. The related dosimetric analysis showed that IR-based patient setup caused limited variations in three cases, with 7% maximum dose reduction in the clinical target volume and no dose increase in organs at risk. CONCLUSIONS: In conclusion, patient setup driven by IR external surrogates localization in 6D featured comparable accuracy with respect to procedures based on stereoscopic X-ray imaging.

3D optoelectronic analysis of interfractional patient setup variability in frameless extracranial stereotactic radiotherapy.

PURPOSE: To investigate size and frequency of interfractional patient setup variability in hypofractionated stereotactic extracranial radiotherapy. METHODS AND MATERIALS: Infrared optical 3D tracking of surface markers was applied to quantify setup variability on 51 patients. Isocenter position repeatability was assessed by means of frameless anatomic calibration and was compared with portal image evaluation. Specific data analysis allowed for compensation of patients' breathing movements and for separation of the effects of operator-dependent misalignments and respiration-induced displacements. Effects of patient position (supine vs. prone) and treatment table configuration were investigated. RESULTS: Patient positioning assisted by the optical tracking device allowed reducing displacements of surface control points within the 3-mm range. Errors in isocenter localization were in the range of a few millimeters. This was in agreement with the portal image evaluation. Breathing motion introduced appreciable errors, which increased control points and isocenter 3D variability. This effect was significantly higher than those related to other investigated factors. CONCLUSIONS: The role of infrared optical tracking devices for patient positioning is assessed on a large patient population. Their use in the frame of high-precision radiotherapy is emphasized by the application of related methodologies for breathing phase detection and frameless isocenter localization.