In radiotherapy, automatic organ-at-risk segmentation algorithms allow faster delineation times, but clinically relevant contour evaluation remains challenging. Commonly used measures to assess automatic contours, such as volumetric Dice Similarity Coefficient (DSC) or Hausdorff distance, have shown to be good measures for geometric similarity, but do not always correlate with clinical applicability of the contours, or time needed to adjust them. This study aimed to evaluate the correlation of new and commonly used evaluation measures with time-saving during contouring. Twenty lung cancer patients were used to compare user-adjustments after atlas-based and deep-learning contouring with manual contouring. The absolute time needed (s) of adjusting the auto-contour compared to manual contouring was recorded, from this relative time-saving (%) was calculated. New evaluation measures (surface DSC and added path length, APL) and conventional evaluation measures (volumetric DSC and Hausdorff distance) were correlated with time-recordings and time-savings, quantified with the Pearson correlation coefficient, R. The highest correlation (R = 0.87) was found between APL and absolute adaption time. Lower correlations were found for APL with relative time-saving (R = −0.38), for surface DSC with absolute adaption time (R = −0.69) and relative time-saving (R = 0.57). Volumetric DSC and Hausdorff distance also showed lower correlation coefficients for absolute adaptation time (R = −0.32 and 0.64, respectively) and relative time-saving (R = 0.44 and −0.64, respectively). Surface DSC and APL are better indicators for contour adaptation time and time-saving when using auto-segmentation and provide more clinically relevant and better quantitative measures for automatically-generated contour quality, compared to commonly-used geometry-based measures.
Proton beam ranges derived from dual-energy computed tomography (DECT) images from a dual-spiral radiotherapy (RT)-specific CT scanner were assessed using Monte Carlo (MC) dose calculations. Images from a dual-source and a twin-beam DECT scanner were also used to establish a comparison to the RT-specific scanner. Proton ranges extracted from conventional single-energy CT (SECT) were additionally performed to benchmark against literature values. Using two phantoms, a DECT methodology was tested as input for G eant 4 MC proton dose calculations. Proton ranges were calculated for different mono-energetic proton beams irradiating both phantoms; the results were compared to the ground truth based on the phantom compositions. The same methodology was applied in a head-and-neck cancer patient using both SECT and dual-spiral DECT scans from the RT-specific scanner. A pencil-beam-scanning plan was designed, which was subsequently optimized by MC dose calculations, and differences in proton range for the different image-based simulations were assessed. For phantoms, the DECT method yielded overall better material segmentation with >86% of the voxel correctly assigned for the dual-spiral and dual-source scanners, but only 64% for a twin-beam scanner. For the calibration phantom, the dual-spiral scanner yielded range errors below 1.2 mm (0.6% of range), like the errors yielded by the dual-source scanner (<1.1 mm, <0.5%). With the validation phantom, the dual-spiral scanner yielded errors below 0.8 mm (0.9%), whereas SECT yielded errors up to 1.6 mm (2%). For the patient case, where the absolute truth was missing, proton range differences between DECT and SECT were on average in −1.2 ± 1.2 mm (−0.5% ± 0.5%). MC dose calculations were successfully performed on DECT images, where the dual-spiral scanner resulted in media segmentation and range accuracy as good as the dual-source CT. In the patient, the various methods showed relevant range differences.
In radiotherapy, computed tomography (CT) datasets are mostly used for radiation treatment planning to achieve a high-conformal tumor coverage while optimally sparing healthy tissue surrounding the tumor, referred to as organs-at-risk (OARs). Based on CT scan and/or magnetic resonance images, OARs have to be manually delineated by clinicians, which is one of the most time-consuming tasks in the clinical workflow. Recent multi-atlas (MA) or deep-learning (DL) based methods aim to improve the clinical routine by an automatic segmentation of OARs on a CT dataset. However, so far no studies investigated the performance of these MA or DL methods on dual-energy CT (DECT) datasets, which have been shown to improve the image quality compared to conventional 120 kVp single-energy CT. In this study, the performance of an in-house developed MA and a DL method (two-step three-dimensional U-net) was quantitatively and qualitatively evaluated on various DECT-derived pseudo-monoenergetic CT datasets ranging from 40 keV to 170 keV. At lower energies, the MA method resulted in more accurate OAR segmentations. Both the qualitative and quantitative metric analysis showed that the DL approach often performed better than the MA method.
To investigate the feasibility of using dual-energy CT (DECT) for tissue segmentation and kilovolt (kV) dose calculations in pre-clinical studies and assess potential dose calculation accuracy gain. Two phantoms and an ex-vivo mouse were scanned in a small animal irradiator with two distinct energies. Tissue segmentation was performed with the single-energy CT (SECT) and DECT methods. A number of different material maps was used. Dose calculations were performed to verify the impact of segmentations on the dose accuracy. DECT showed better overall results in comparison to SECT. Higher number of DECT segmentation media resulted in smaller dose differences in comparison to the reference. Increasing the number of materials in the SECT method yielded more instability. Both modalities showed a limit to which adding more materials with similar characteristics ceased providing better segmentation results, and resulted in more noise in the material maps and the dose distributions. The effect was aggravated with a decrease in beam energy. For the ex-vivo specimen, the choice of only one high dense bone for the SECT method resulted in large volumes of tissue receiving high doses. For the DECT method, the choice of more than one kind of bone resulted in lower dose values for the different tissues occupying the same volume. For the organs at risk surrounded by bone, the doses were lower when using the SECT method in comparison to DECT, due to the high absorption of the bone. SECT material segmentation may lead to an underestimation of the dose to OAR in the proximity of bone. The DECT method enabled the selection of a higher number of materials thereby increasing the accuracy in dose calculations. In phantom studies, SECT performed best with three materials and DECT with seven for the phantom case. For irradiations in preclinical studies with kV photon energies, the use of DECT segmentation combined with the choice of a low-density bone is recommended.
1 March 2017 in Physics and Imaging in Radiation Oncology
Background and purpose: Computed tomography (CT) imaging is frequently used in radiation oncology to calculate radiation dose distributions. In order to calculate doses, the CT numbers must be converted into densities by an energy dependent conversion curve. A recently developed algorithm directly reconstructs CT projection data into relative electron densities which eliminates the use of separate conversion curves for different X-ray tube potentials. Our work evaluates this algorithm for various cancer sites and shows its applicability in a clinical workflow. Materials and methods: The Gammex phantom with tissue mimicking inserts was scanned to characterize the CT number to density conversion curves. In total, 33 patients with various cancer sites were scanned using multiple tube potentials. All CT acquisitions were reconstructed with the standard filtered back-projection (FBP) and the new developed DirectDensity⢠(DD) algorithm. The mean tumor doses and the volume percentage that receives more than 95% of the prescribed dose were calculated for the planning target volume. Relevant parameters for the organs at risk for each tumor site were also calculated. Results: The relative mean dose differences between the standard 120 kVp FBP CT scan workflow and the DD CT scans (80, 100, 120 and 140 kVp) were in general less than 1% for the planned target volume and organs at risk. Conclusion: The energy independent DD algorithm allows for accurate dose calculations over a variety of body sites. This novel algorithm eliminates the tube potential specific calibration procedure and thereby simplifies the clinical radiotherapy workflow. Keywords: CT imaging, Image reconstruction, Dose calculations, Electron density reconstructio
About Me
Postdoctoral scientist with research interests in medical physics and computer vision.
Postdoctoral scientist with research interests in medical physics and computer vision.