Présent et avenir de la radiothérapie guidée par l'image (IGRT) et ses applications possibles dans le traitement des cancers bronchiques

Cone-beam CT image quality improvement using Cycle-Deblur consistent adversarial networks (Cycle-Deblur GAN) for chest CT imaging in breast cancer patients

Cone-beam computed tomography (CBCT) integrated with a linear accelerator is widely used to increase the accuracy of radiotherapy and plays an important role in image-guided radiotherapy (IGRT). For comparison with fan-beam computed tomography (FBCT), the image quality of CBCT is indistinct due to X-ray scattering, noise, and artefacts. We proposed a deep learning model, “Cycle-Deblur GAN”, combined with CycleGAN and Deblur-GAN models to improve the image quality of chest CBCT images. The 8706 CBCT and FBCT image pairs were used for training, and 1150 image pairs were used for testing in deep learning. The generated CBCT images from the Cycle-Deblur GAN model demonstrated closer CT values to FBCT in the lung, breast, mediastinum, and sternum compared to the CycleGAN and RED-CNN models. The quantitative evaluations of MAE, PSNR, and SSIM for CBCT generated from the Cycle-Deblur GAN model demonstrated better results than the CycleGAN and RED-CNN models. The Cycle-Deblur GAN model improved image quality and CT-value accuracy and preserved structural details for chest CBCT images.

Reduced toxicity with three-dimensional conformal radiotherapy or intensity-modulated radiotherapy compared with conventional two-dimensional radiotherapy for esophageal squamous cell carcinoma: a secondary analysis of data from four prospective clinical trials

We conducted a retrospective analysis to assess the toxicity and long-term survival of esophageal squamous cell carcinoma patients treated with three-dimensional conformal radiotherapy (3DCRT) or intensity-modulated radiotherapy (IMRT) versus conventional two-dimensional radiotherapy (2DRT). All data in the present study were based on four prospective clinical trials conducted at our institution from 1996 to 2004 and included 308 esophageal squamous cell carcinoma patients treated with 2DRT or 3DCRT/IMRT. Based on the inclusion and exclusion criteria, 254 patients were included in the analysis. Of these patients, 158 were treated with 2DRT, whereas 96 were treated with 3DCRT/IMRT. The rates of ≥Grade3 acute toxicity of the esophagus and lung were 11.5% versus 28.5% (P = 0.002) and 5.2% versus 10.8% (P = 0.127) in the 3DCRT/IMRT and 2DRT groups, respectively. The incidences of ≥Grade 3 late toxicity of the esophagus and lungs were 3.1% versus 10.7% (P = 0.028) and 3.1% versus 5.7% (P = 0.127) in the 3DCRT/IMRT and 2DRT groups, respectively. The 1-year, 3-year and 5-year estimated overall survival rates were 81%, 38% and 34% in the 3DCRT/IMRT group and 79%, 44% and 31% in the 2DRT group, respectively (P = 0.628). The 1-year, 3-year and 5-year local control rates were 88%, 71% and 66% in the 3DCRT/IMRT group and 84%, 66% and 60% in the 2DRT group, respectively (P = 0.412). Fewer incidences of acute and late toxicities were observed in esophageal squamous cell carcinoma patients treated with 3DCRT/IMRT compared with those treated with 2DRT. No significant survival benefit was observed with the use of 3DCRT/IMRT.

A software tool of digital tomosynthesis application for patient positioning in radiotherapy

Digital Tomosynthesis (DTS) is an image modality in reconstructing tomographic images from two‐dimensional kV projections covering a narrow scan angles. Comparing with conventional cone‐beam CT (CBCT), it requires less time and radiation dose in data acquisition. It is feasible to apply this technique in patient positioning in radiotherapy. To facilitate its clinical application, a software tool was developed and the reconstruction processes were accelerated by graphic processing unit (GPU). Two reconstruction and two registration processes are required for DTS application which is different from conventional CBCT application which requires one image reconstruction process and one image registration process. The reconstruction stage consists of productions of two types of DTS. One type of DTS is reconstructed from cone‐beam (CB) projections covering a narrow scan angle and is named onboard DTS (ODTS), which represents the real patient position in treatment room. Another type of DTS is reconstructed from digitally reconstructed radiography (DRR) and is named reference DTS (RDTS), which represents the ideal patient position in treatment room. Prior to the reconstruction of RDTS, The DRRs are reconstructed from planning CT using the same acquisition setting of CB projections. The registration stage consists of two matching processes between ODTS and RDTS. The target shift in lateral and longitudinal axes are obtained from the matching between ODTS and RDTS in coronal view, while the target shift in longitudinal and vertical axes are obtained from the matching between ODTS and RDTS in sagittal view. In this software, both DRR and DTS reconstruction algorithms were implemented on GPU environments for acceleration purpose. The comprehensive evaluation of this software tool was performed including geometric accuracy, image quality, registration accuracy, and reconstruction efficiency. The average correlation coefficient between DRR/DTS generated by GPU‐based algorithm and CPU‐based algorithm is 0.99. Based on the measurements of cube phantom on DTS, the geometric errors are within 0.5 mm in three axes. For both cube phantom and pelvic phantom, the registration errors are within 0.5 mm in three axes. Compared with reconstruction performance of CPU‐based algorithms, the performances of DRR and DTS reconstructions are improved by a factor of 15 to 20. A GPU‐based software tool was developed for DTS application for patient positioning of radiotherapy. The geometric and registration accuracy met the clinical requirement in patient setup of radiotherapy. The high performance of DRR and DTS reconstruction algorithms was achieved by the GPU‐based computation environments. It is a useful software tool for researcher and clinician in evaluating DTS application in patient positioning of radiotherapy.

PACS number(s): 87.57.nf

Application of an independent dose calculation software for estimating the impact of inter-fractional setup shifts in Helical Tomotherapy treatments

The purpose of this study is to validate the capability of in-house independent point dose calculation software to be used as a second check for Helical Tomotherapy treatment plans. The software performed its calculations in homogenous conditions (using the Cheese phantom, which is a cylindrical phantom with radius 15 cm and length 18 cm) using a factor-based algorithm. Fifty patients, who were treated for pelvic (10), prostate (14), lung (10), head & neck (12) and brain (4) cancers, were used. Based on the individual patient kVCT images and the pretreatment MVCT images for each treatment fraction, the corresponding daily patient setup shifts in the IEC-X, IEC-Y, and IEC-Z directions were registered. For each patient, the registered fractional setup shifts were grouped into systematic and random shifts. The average systematic dosimetric variations showed small dose deviation for the different cancer types (1.0%-3.0%) compared to the planned dose. Of the fifty patients, only three had percent differences larger than 5%. The average random dosimetric variations showed relatively small dose deviations (0.2%-1.1%) compared to the planned dose. None of the patients had percent differences larger than 5%. By examining the individual fractions of each patient, it is observed that only in 31 out of 1358 fractions the percent differences exceeded the border of 5%. These results indicate that the overall dosimetric impact from systematic and random variations is small and that the software is a capable platform for independent point dose validation for the Helical Tomotherapy modality.

Three-dimensional patient setup errors at different treatment sites measured by the Tomotherapy megavoltage CT

BACKGROUND AND PURPOSE

Reduction of interfraction setup uncertainty is vital for assuring the accuracy of conformal radiotherapy. We report a systematic study of setup error to assess patients' three-dimensional (3D) localization at various treatment sites.

PATIENTS AND METHODS

Tomotherapy megavoltage CT (MVCT) images were scanned daily in 259 patients from 2005-2008. We analyzed 6,465 MVCT images to measure setup error for head and neck (H&N), chest/thorax, abdomen, prostate, legs, and total marrow irradiation (TMI). Statistical comparisons of the absolute displacements across sites and time were performed in rotation (R), lateral (x), craniocaudal (y), and vertical (z) directions.

RESULTS

The global systematic errors were measured to be less than 3 mm in each direction with increasing order of errors for different sites: H&N, prostate, chest, pelvis, spine, legs, and TMI. The differences in displacements in the x, y, and z directions, and 3D average displacement between treatment sites were significant (p < 0.01). Overall improvement in patient localization with time (after 3-4 treatment fractions) was observed. Large displacement (> 5 mm) was observed in the 75(th) percentile of the patient groups for chest, pelvis, legs, and spine in the x and y direction in the second week of the treatment.

CONCLUSION

MVCT imaging is essential for determining 3D setup error and to reduce uncertainty in localization at all anatomical locations. Setup error evaluation should be performed daily for all treatment regions, preferably for all treatment fractions.

Comparison of IGRT registration strategies for optimal coverage of primary lung tumors and involved nodes based on multiple four-dimensional CT scans obtained throughout the radiotherapy course

PURPOSE

To investigate the impact of primary tumor and involved lymph node (LN) geometry (centroid, shape, volume) on internal target volume (ITV) throughout treatment for locally advanced non-small cell lung cancer using weekly four-dimensional computed tomography (4DCT).

METHODS AND MATERIALS

Eleven patients with advanced non-small cell lung cancer were treated using image-guided radiotherapy with acquisition of weekly 10-Phase 4DCTs (n = 51). Initial ITV was based on planning 4DCT. Master-ITV incorporated target geometry across the entire treatment (all 4DCTs). Geographic miss was defined as the % Master-ITV positioned outside of the initial planning ITV after registration is complete. Registration strategies considered were bony (B), primary tumor soft tissue alone (T), and registration based on primary tumor and involved LNs (T_LN).

RESULTS

The % geographic miss for the primary tumor, mediastinal, and hilar lymph nodes based on each registration strategy were (1) B: 30%, 30%, 30%; (2) T: 21%, 40%, 36%; and (3) T_LN: 26%, 26%, 27%. Mean geographic expansions to encompass 100% of the primary tumor and involved LNs were 1.2 ± 0.7 cm and 0.8 ± 0.3 cm, respectively, for B and T_LN. Primary and involved LN expansions were 0.7 ± 0.5 cm and 1.1 ± 0.5 cm for T.

CONCLUSION

T is best for solitary targets. When treatments include primary tumor and LNs, B and T_LN provide more comprehensive geographic coverage. We have identified high % geographic miss when considering multiple registration strategies. The dosimetric implications are the subject of future study.

DART, a platform for the creation and registration of cone beam digital tomosynthesis datasets

Digital tomosynthesis is an imaging modality that allows for tomographic reconstructions using only a fraction of the images needed for CT reconstruction. Since it offers the advantages of tomographic images with a smaller imaging dose delivered to the patient, the technique offers much promise for use in patient positioning prior to radiation delivery. This paper describes a software environment developed to help in the creation of digital tomosynthesis image sets from digital portal images using three different reconstruction algorithms. The software then allows for use of the tomograms for patient positioning or for dose recalculation if shifts are not applied, possibly as part of an adaptive radiotherapy regimen.

The effect of a limited number of projections and reconstruction algorithms on the image quality of megavoltage digital tomosynthesis

In order to investigate the effect of the number of projections on digital tomosynthesis image quality, images were acquired over a 40 degree arc and sampled into sets of 2 to 41 projections used as input to three different reconstruction algorithms: the shift‐and‐add, the Feldkamp‐Davis‐Kress filtered back projection algorithms, and the simultaneous algebraic reconstruction technique. The variation of several image characteristics, such as in‐plane resolution, contrast to noise ratio, artifact spread, volumetric accuracy, and dose, are investigated based on the reconstruction algorithms used and also the number of projections used as source data. The results suggest that only 11 projections are required since the various parameters checked do not improve much past that number. As a reconstruction algorithm, SART did best but took much longer to reconstruct images. Thus, if reconstruction time is a determining factor, filtered back‐projection looks like a better compromise.

PACS number: 87.57.C‐, 87.57.nf, 87.57.Q‐, 87.59.‐e

[Quality assurance of a virtual simulation software: application to IMAgo and SIMAgo (ISOgray)]

PURPOSE

Virtual simulation process is often used to prepare three dimensional conformal radiation therapy treatments. As the quality of the treatment is widely dependent on this step, it is mandatory to perform extensive controls on this software before clinical use. The tests presented in this work have been carried out on the treatment planning system ISOgray (DOSIsoft), including the delineation module IMAgo and the virtual simulation module SIMAgo.

MATERIAL AND METHODS

According to our experience, the most relevant controls of international protocols have been selected. These tests mainly focused on measuring and delineation tools, virtual simulation functionalities, and have been performed with three phantoms: the Quasar Multi-Purpose Body Phantom, the Quasar MLC Beam Geometry Phantom (Modus Medical Devices Inc.) and a phantom developed at Hospital Tenon.

RESULTS

No major issues have been identified while performing the tests. These controls have emphasized the necessity for the user to consider with a critical eye the results displayed by a virtual simulation software. The contrast of visualisation, the slice thickness, the calculation and display mode of 3D structures used by the software are many factors of uncertainties.

CONCLUSION

A virtual simulation software quality assurance procedure has been written and applied on a set of CT images. Similar tests have to be performed periodically and at minimum at each change of major version.