Search

Background: Automatic segmentation of vertebrae in spinal x-ray images is crucial for clinical diagnosis, case analysis, and surgical planning of spinal lesions., Purpose: However, due to the inherent characteristics of x-ray images, including low contrast, high noise, and uneven grey scale, it remains a critical and challenging problem in computer-aided spine image analysis and disease diagnosis applications., Methods: In this paper, a Multiscale Feature Enhancement Network (MFENet), is proposed for segmenting whole spinal x-ray images, to aid doctors in diagnosing spinal-related diseases. To enhance feature extraction, the network incorporates a Dual-branch Feature Extraction Module (DFEM) and a Semantic Aggregation Module (SAM). The DFEM has a parallel dual-branch structure. The upper branch utilizes multiscale convolutional kernels to extract features from images. Employing convolutional kernels of different sizes helps capture details and structural information at different scales. The lower branch incorporates attention mechanisms to further optimize feature representation. By modeling the feature maps spatially and across channels, the network becomes more focused on key feature regions and suppresses task-irrelevant information. The SAM leverages contextual semantic information to compensate for details lost during pooling and convolution operations. It integrates high-level feature information from different scales to reduce segmentation result discontinuity. In addition, a hybrid loss function is employed to enhance the network's feature extraction capability., Results: In this study, we conducted a multitude of experiments utilizing dataset provided by the Spine Surgery Department of Henan Provincial People's Hospital. The experimental results indicate that our proposed MFENet demonstrates superior segmentation performance in spinal segmentation on x-ray images compared to other advanced methods, achieving 92.61 ± 0.431 for MIoU, 92.42 ± 0.329 for DSC, and 99.51 ± 0.037 for Global_accuracy., Conclusions: Our model is able to more effectively learn and extract global contextual semantic information, significantly improving spinal segmentation performance, further aiding doctors in analyzing patient conditions., (© 2024 American Association of Physicists in Medicine.)

Background: Accurate and automated segmentation of thoracic organs-at-risk (OARs) is critical for radiotherapy treatment planning of thoracic cancers. However, this has remained a challenging task for four major reasons: (1) thoracic OARs have diverse morphologies; (2) thoracic OARs have low contrast with the background; (3) boundaries of thoracic OARs are blurry; (4) class imbalance issue caused by small organs., Purpose: To overcome the above challenges and achieve accurate and automated segmentation of thoracic OARs on thoracic CT., Methods: A novel cascaded framework based on mixed attention and multiscale information for thoracic OARs segmentation, called Cascaded-TOARNet. This cascaded framework comprises two stages: localization and segmentation. During the localization stage, TOARNet locates each organ to crop the regions of interest (ROIs). During the segmentation stage, TOARNet accurately segments the ROIs, and the segmentation results are merged into a complete result., Results: We evaluated our proposed method and other common segmentation methods on two public datasets: the AAPM Thoracic Auto-Segmentation Challenge dataset and the Segmentation of Thoracic Organs at Risk (SegTHOR) dataset. Our method demonstrated superior performance, achieving a mean Dice score of 92.6% on the SegTHOR dataset and 90.8% on the AAPM dataset., Conclusions: This segmentation method holds great promise as an essential tool for enhancing the efficiency of thoracic radiotherapy planning., (© 2023 American Association of Physicists in Medicine.)

Purpose: Excessive complexity in intensity-modulated radiation therapy (IMRT) plans increases the dose uncertainty, prolongs the treatment time, and increases the susceptibility to changes in patient or target geometry. To date, the tools for quantitative assessment of IMRT beam complexity are still lacking. In this study, The authors have sought to develop metrics to characterize different aspects of beam complexity and investigate the beam complexity for IMRT plans of different disease sites., Methods: The authors evaluated the beam complexity scores for 65 step-and-shoot IMRT plans from three sites (prostate, head and neck, and spine) and 26 volumetric-modulated arc therapy (VMAT) plans for the prostate. On the basis of the beam apertures and monitor unit weights of all segments, the authors calculated the mean aperture area, extent of aperture shape irregularity, and degree of beam modulation for each beam. Then the beam complexity values were averaged to obtain the complexity metrics of the IMRT plans. The authors studied the correlation between the beam complexity metrics and the quality assurance (QA) results. Finally, the effects of treatment planning parameters on beam complexity were studied., Results: The beam complexity scores were not uniform among the prostate IMRT beams from different gantry angles. The lateral beams had larger monitor units and smaller shape irregularity, while the anterior-posterior beams had larger modulation values. On average, the prostate IMRT plans had the smallest aperture irregularity, beam modulation, and normalized monitor units; the head and neck IMRT plans had large beam irregularity and beam modulation; and the spine stereotactic radiation therapy plans often had small beam apertures, which may have been associated with the relatively large discrepancies between planned and QA measured doses. There were weak correlations between the beam complexity scores and the measured dose errors. The prostate VMAT beams showed greater complexity than the prostate step-and-shoot IMRT beams. In the treatment planning process, the beam complexity increased as the minimum segment area decreased and as the number of optimization iterations and the maximum number of segments increased., Conclusions: The proposed metrics were effective in characterizing the beam complexity of different disease sites and for different treatment modalities. Efforts should be made to reduce the unnecessary complexity of IMRT beams to minimize the radiation dose uncertainties.

Purpose: Gantry sag is one of the well-known sources of mechanical imperfections that compromise the spatial accuracy of radiation dose delivery. The objectives of this study were to quantify the gantry sag on multiple linear accelerators (linacs), to investigate a multileaf collimator (MLC)-based strategy to compensate for gantry sag, and to verify the gantry sag and its compensation with film measurements., Methods: The authors used the Winston-Lutz method to measure gantry sag on three Varian linacs. A ball bearing phantom was imaged with megavolt radiation fields at 10° gantry angle intervals. The images recorded with an electronic portal imaging device were analyzed to derive the radiation isocenter and the gantry sag, that is, the superior-inferior wobble of the radiation field center, as a function of the gantry angle. The authors then attempted to compensate for the gantry sag by applying a gantry angle-specific correction to the MLC leaf positions. The gantry sag and its compensation were independently verified using film measurements., Results: Gantry sag was reproducible over a six-month measurement period. The maximum gantry sag was found to vary from 0.7 to 1.0 mm, depending on the linac and the collimator angle. The radiation field center moved inferiorly (i.e., away from the gantry) when the gantry was rotated from 0° to 180°. After the MLC leaf position compensation was applied at 90° collimator angle, the maximum gantry sag was reduced to <0.2 mm. The film measurements at gantry angles of 0° and 180° verified the inferior shift of the radiation fields and the effectiveness of MLC compensation., Conclusions: The results indicate that gantry sag on a linac can be quantitatively measured using a simple phantom and an electronic portal imaging device. Reduction of gantry sag is feasible by applying a gantry angle-specific correction to MLC leaf positions at 90° collimator angle.

Purpose: The isocenter accuracy of a linear accelerator is often assessed with star-shot films. This approach is limited in its ability to quantify three dimensional wobble of radiation field centers (RFCs). The authors report a Winston-Lutz based method to measure the 3D wobble of RFCs during gantry rotation, collimator rotation, and collimator field size change., Methods: A stationary ball-bearing phantom was imaged using multileaf collimator-shaped radiation fields at various gantry angles, collimator angles, and field sizes. The center of the ball-bearing served as a reference point, to which all RFCs were localized using a computer algorithm with sub-pixel accuracy. Then, the gantry rotation isocenter and the collimator rotation axis were derived from the coordinates of these RFCs. Finally, the deviation or wobble of the individual RFC from the derived isocenter or rotation axis was quantified., Results: The results showed that the RFCs were stable as the field size of the multileaf collimator was varied. The wobble of RFCs depended on the gantry angle and the collimator angle and was reproducible, indicating that the mechanical imperfections of the linac were mostly systematic and quantifiable. It was found that the 3D wobble of RFCs during gantry rotation was reduced after compensating for a constant misalignment of the multileaf collimator., Conclusions: The 3D wobble of RFCs can be measured with submillimeter precision using the proposed method. This method provides a useful tool for checking and adjusting the radiation isocenter tightness of a linac.

Purpose: The aim of this study was to develop a computerized method to quantify the coincidence between portal image graticules and radiation field centers or radiation isocenter. Three types of graticules were included in this study: Megavoltage (MV) mechanical graticule, MV electronic portal imaging device digital graticule, and kilovoltage (kV) on-board imaging digital graticule., Methods: A metal ball bearing (BB) was imaged with MV and kV x-ray beams in a procedure similar to a Winston-Lutz test. The radiation fields, graticules, and BB were localized in eight portal images using Hough transform-based computer algorithms. The center of the BB served as a static reference point in the 3D space so that the distances between the graticule centers and the radiation field centers were calculated. The radiation isocenter was determined from the radiation field centers at different gantry angles., Results: Misalignments of MV and kV portal imaging graticules varied with the gantry or x-ray source angle as a result of mechanical imperfections of the linear accelerator and its imaging system. While the three graticules in this study were aligned to the radiation field centers and the radiation isocenter within 2.0 mm, misalignments of 1.5-2.0 mm were found at certain gantry angles. These misalignments were highly reproducible with the gantry rotation., Conclusions: A simple method was developed to quantify the alignments of portal image graticules directly against the radiation field centers or the radiation isocenter. The advantage of this method is that it does not require the BB to be placed exactly at the radiation isocenter through a precalibrated surrogating device such as room lasers or light field crosshairs. The present method is useful for radiation therapy modalities that require high-precision portal imaging such as image-guided stereotactic radiotherapy.

Magnetic resonance images can be compromised by constant frequency (CF) noise, such as radio frequency (RF) noise. We investigate the effects of CF noise in four imaging methods with non-uniform k-space sampling: single-shot echo-planar imaging (EPI) with sinusoidal readout gradients, phase-encoded echo-planar spectroscopic imaging (EPSI) with sinusoidal readout gradients, projection-reconstruction imaging, and single-shot spiral imaging. The characteristics of the CF artifacts in each imaging method are studied with numerical simulations. CF noise is found to cause artifacts of nonclassic forms in the reconstructed images. Specifically, dashed-line, streak, and spiral patterns of CF noise appear in EPI/EPSI with sinusoidal readout gradients, projection-reconstruction imaging, and spiral imaging, respectively. The analytical expression for dashed-line artifacts is found to be a weighted sum of Bessel functions and is verified with in vivo experiments. The CF artifacts can be removed during post-processing by eliminating the noise spikes in the Fourier domain of the raw data.