3D CT Images Research Articles

Interplay Effect of Splenic Motion for Total Lymphoid Irradiation in Pediatric Proton Therapy.

(1) Background: The most significant cause of an unacceptable deviation from the planned dose during respiratory motion is the interplay effect. We examined the correlation between the magnitude of splenic motion and its impact on plan quality for total lymphoid irradiation (TLI); (2) Methods: Static and 4D CT images from ten patients were used for interplay effect simulations. Patients' original plans were optimized based on the average CT extracted from the 4D CT and planned with two posterior beams using scenario-based optimization (±3 mm of setup and ±3% of range uncertainty) and gradient matching at the level of mid-spleen. Dynamically accumulated 4D doses (interplay effect dose) were calculated based on the time-dependent delivery sequence of radiation fluence across all phases of the 4D CT. Dose volume parameters for each simulated treatment delivery were evaluated for plan quality; (3) Results: Peak-to-peak splenic motion (≤12 mm) was measured from the 4D CT of ten patients. Interplay effect simulations revealed that the ITV coverage of the spleen remained within the protocol tolerance for splenic motion, ≤8 mm. The D100% coverage for ITV spleen decreased from 95.0% (nominal plan) to 89.3% with 10 mm and 87.2% with 12 mm of splenic motion; (4) Conclusions: 4D plan evaluation and robust optimization may overcome problems associated with respiratory motion in proton TLI treatments. Patient-specific respiratory motion evaluations are essential to confirming adequate dosimetric coverage when proton therapy is utilized.

Applicability of postmortem CT organ-selective volume rendering in evaluating liver and heart injuries

ObjectivesPostmortem computed tomography (PMCT) is widely used in forensic cause-of-death investigations, and 3D volume rendering is commonly employed for visualizing PMCT data. However, no studies have investigated the effectiveness of organ-selective 3D volume rendering for detection of organ injuries such as liver and heart. This study aimed to assess whether selective volume rendering could enhance the detectability of organ damage that is generally difficult to observe in 2DCT images. Material and methodsUsing 22 selected cases of liver or heart injury, the detectability of the injury in 2DCT images and selective volume-rendered images was compared with actual autopsy findings. ResultsAmong 15 cases of liver injury confirmed at autopsy, 2DCT imaging detected two, whereas selective volume-rendered imaging detected one. Furthermore, among 17 cases of cardiac injury, 2DCT imaging detected two, whereas selective volume-rendered imaging detected four. Thus, a small proportion of liver and heart injury cases were detectable using 2DCT imaging or selective volume rendering. In all cases where selective volume rendering identified an injury, the injury was not identified by radiologist interpretation of the 2DCT images, although a review of the 2DCT images did allow identification of the injury when corroborated by the findings of selective volume rendering. One additional case demonstrated the usefulness of selective volume rendering for estimation of puncture wound direction. ConclusionsAlthough selective volume rendering has seldom been effective in detecting heart or liver damage, organ-selective volume rendering may provide clues for detection of injuries that are often overlooked in 2DCT images alone.

Dose reduction in sequence scanning 4D CT imaging through respiratory signal-guided tube current modulation: A feasibility study.

Respiratory signal-guided 4D CT sequence scanning such as the recently introduced Intelligent 4D CT (i4DCT) approach reduces image artifacts compared to conventional 4D CT, especially for irregular breathing. i4DCT selects beam-on periods during scanning such that data sufficiency conditions are fulfilled for each couch position. However, covering entire breathing cycles during beam-on periods leads to redundant projection data and unnecessary dose to the patient during long exhalation phases. We propose and evaluate the feasibility of respiratory signal-guided dose modulation (i.e., temporary reduction of the CT tube current) to reduce the i4DCT imaging dose while maintaining high projection data coverage for image reconstruction. The study is designed as an in-silico feasibility study. Dose down- and up-regulation criteria were defined based on the patients' breathing signals and their representative breathing cycle learned before and during scanning. The evaluation (including an analysis of the impact of the dose modulation criteria parameters) was based on 510 clinical 4D CT breathing curves. Dose reduction was determined as the fraction of the downregulated dose delivery time to the overall beam-on time. Furthermore, under the assumption of a 10-phase 4D CT and amplitude-based reconstruction, beam-on periods were considered negatively affected by dose modulation if the downregulation period covered an entire phase-specific amplitude range for a specific breathing phase (i.e., no appropriate reconstruction of the phase image possible for this specific beam-on period). Corresponding phase-specific amplitude bins are subsequently denoted as compromised bins. Dose modulation resulted in a median dose reduction of 10.4% (lower quartile: 7.4%, upper quartile: 13.8%, maximum: 28.6%; all values corresponding to a default parameterization of the dose modulation criteria). Compromised bins were observed in 1.0% of the beam-on periods (72 / 7370 periods) and affected 10.6% of the curves (54/510 curves). The extent of possible dose modulation depends strongly on the individual breathing patterns and is weakly correlated with the median breathing cycle length (Spearman correlation coefficient 0.22, p<0.001). Moreover, the fraction of beam-on periods with compromised bins is weakly anti-correlated with the patient's median breathing cycle length (Spearman correlation coefficient -0.24; p<0.001). Among the curves with the 17% longest average breathing cycles, no negatively affected beam-on periods were observed. Respiratory signal-guided dose modulation for i4DCT imaging is feasible and promises to significantly reduce the imaging dose with little impact on projection data coverage. However, the impact on image quality remains to be investigated in a follow-up study.

Comparative evaluation of a surface-based respiratory monitoring system against a pressure sensor for 4DCT image reconstruction in phantoms.

Four-dimensional computed tomography (4DCT), which relies on breathing-induced motion, requires realistic surrogate information of breathing variations to reconstruct the tumor trajectory and motion variability of normal tissues accurately. Therefore, the SimRT surface-guided respiratory monitoring system has been installed on a Siemens CT scanner. This work evaluated the temporal and spatial accuracy of SimRT versus our commonly used pressure sensor, AZ-733V. A dynamic thorax phantom was used to reproduce regular and irregular breathing patterns acquired by SimRT and Anzai. Various parameters of the recorded breathing patterns, including mean absolute deviations (MAD), Pearson correlations (PC), and tagging precision, were investigated and compared to ground-truth. Furthermore, 4DCT reconstructions were analyzed to assess the volume discrepancy, shape deformation and tumor trajectory. Compared to the ground-truth, SimRT more precisely reproduced the breathing patterns with a MAD range of 0.37±0.27 and 0.92±1.02mm versus Anzai with 1.75±1.54 and 5.85±3.61mm for regular and irregular breathing patterns, respectively. Additionally, SimRT provided a more robust PC of 0.994±0.009 and 0.936±0.062 for all investigated breathing patterns. Further, the peak and valley recognition were found to be more accurate and stable using SimRT. The comparison of tumor trajectories revealed discrepancies up to 7.2 and 2.3mm for Anzai and SimRT, respectively. Moreover, volume discrepancies up to 1.71±1.62% and 1.24±2.02% were found for both Anzai and SimRT, respectively. SimRT was validated across various breathing patterns and showed a more precise and stable breathing tracking, (i) independent of the amplitude and period, (ii) and without placing any physical devices on the patient's body. These findings resulted in a more accurate temporal and spatial accuracy, thus leading to a more realistic 4DCT reconstruction and breathing-adapted treatment planning.

Comparison and evaluation of different deep learning models of synthetic CT generation from CBCT for nasopharynx cancer adaptive proton therapy.

Cone-beam computed tomography (CBCT) scanning is used for patient setup in image-guided radiotherapy. However, its inaccurate CT numbers limit its applicability in dose calculation and treatment planning. This study compares four deep learning methods for generating synthetic CT (sCT) to determine which method is more appropriate and offers potential for further clinical exploration in adaptive proton therapy for nasopharynx cancer. CBCTs and deformed planning CT (dCT) from 75 patients (60/5/10 for training, validation and testing) were used to compare cycle-consistent Generative Adversarial Network(cycleGAN), Unet, Unet+cycleGAN and conditionalGenerative Adversarial Network (cGAN) for sCT generation. The sCT images generated by each method were evaluated against dCT images using mean absolute error (MAE), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), spatial non-uniformity (SNU) and radial averaging in the frequency domain. In addition, dosimetric accuracy was assessed through gamma analysis, differences in water equivalent thickness (WET), and dose-volume histogram metrics. The cGAN model has demonstrated optimal performance in the four models across various indicators. In terms of image quality under global condition, the average MAE has been reduced to 16.39HU, SSIM has increased to 95.24%, and PSNR has increased to 28.98. Regarding dosimetric accuracy, the gamma passing rate (2%/2mm) has reached 99.02%, and the WET difference is only 1.28mm. The D95 value of CTVs coverage and Dmax value of spinal cord, brainstem show no significant differences between dCT and sCT generated by cGAN model. The cGAN model has been shown to be a more suitable approach for generating sCT using CBCT, considering its characteristics and concepts. The resulting sCT has the potential for application in adaptive proton therapy.

Transfer learning from T1‐weighted to T2‐weighted Magnetic resonance sequences for brain image segmentation

AbstractMagnetic resonance (MR) imaging is a widely employed medical imaging technique that produces detailed anatomical images of the human body. The segmentation of MR images plays a crucial role in medical image analysis, as it enables accurate diagnosis, treatment planning, and monitoring of various diseases and conditions. Due to the lack of sufficient medical images, it is challenging to achieve an accurate segmentation, especially with the application of deep learning networks. The aim of this work is to study transfer learning from T1‐weighted (T1‐w) to T2‐weighted (T2‐w) MR sequences to enhance bone segmentation with minimal required computation resources. With the use of an excitation‐based convolutional neural networks, four transfer learning mechanisms are proposed: transfer learning without fine tuning, open fine tuning, conservative fine tuning, and hybrid transfer learning. Moreover, a multi‐parametric segmentation model is proposed using T2‐w MR as an intensity‐based augmentation technique. The novelty of this work emerges in the hybrid transfer learning approach that overcomes the overfitting issue and preserves the features of both modalities with minimal computation time and resources. The segmentation results are evaluated using 14 clinical 3D brain MR and CT images. The results reveal that hybrid transfer learning is superior for bone segmentation in terms of performance and computation time with DSCs of 0.5393 ± 0.0007. Although T2‐w‐based augmentation has no significant impact on the performance of T1‐w MR segmentation, it helps in improving T2‐w MR segmentation and developing a multi‐sequences segmentation model.

MisMatch: Calibrated Segmentation via Consistency on Differential Morphological Feature Perturbations With Limited Labels.

Semi-supervised learning (SSL) is a promising machine learning paradigm to address the ubiquitous issue of label scarcity in medical imaging. The state-of-the-art SSL methods in image classification utilise consistency regularisation to learn unlabelled predictions which are invariant to input level perturbations. However, image level perturbations violate the cluster assumption in the setting of segmentation. Moreover, existing image level perturbations are hand-crafted which could be sub-optimal. In this paper, we propose MisMatch, a semi-supervised segmentation framework based on the consistency between paired predictions which are derived from two differently learnt morphological feature perturbations. MisMatch consists of an encoder and two decoders. One decoder learns positive attention for foreground on unlabelled data thereby generating dilated features of foreground. The other decoder learns negative attention for foreground on the same unlabelled data thereby generating eroded features of foreground. We normalise the paired predictions of the decoders, along the batch dimension. A consistency regularisation is then applied between the normalised paired predictions of the decoders. We evaluate MisMatch on four different tasks. Firstly, we develop a 2D U-net based MisMatch framework and perform extensive cross-validation on a CT-based pulmonary vessel segmentation task and show that MisMatch statistically outperforms state-of-the-art semi-supervised methods. Secondly, we show that 2D MisMatch outperforms state-of-the-art methods on an MRI-based brain tumour segmentation task. We then further confirm that 3D V-net based MisMatch outperforms its 3D counterpart based on consistency regularisation with input level perturbations, on two different tasks including, left atrium segmentation from 3D CT images and whole brain tumour segmentation from 3D MRI images. Lastly, we find that the performance improvement of MisMatch over the baseline might originate from its better calibration. This also implies that our proposed AI system makes safer decisions than the previous methods.

CT-on-Rails Utilization for Image Guidance and Plan Adaptation at a Single-Room Proton Therapy Center.

In-room CT-on-rails is a relatively rare imaging modality in proton therapy. This work evaluated the utilization of CT-on-rails (CTOR) for image guidance and plan adaptation at a single-room proton therapy center. All patients treated from the proton center beginning treatment to the current date were evaluated (May 2021 - January 2023). Patients were treated using a Mevion S250i scanning beam proton therapy system with an in-room Definition Edge CTOR. The image guidance technique utilized for each patient was evaluated (CTOR, orthogonal planar kV imaging, or both). For patients who required a new CT for plan adaptation, the use of CT images from CTOR or a new simulation CT was assessed. Of the 132 patients treated over the evaluated time period, 56.0% utilized CTOR for image guidance at some point during their treatment. 21.2% utilized CTOR as the only image guidance technique, while 34.8% utilized CTOR in combination with kV planar imaging. For 43.9% of patients, only orthogonal planar kV imaging was utilized. Head and neck treatment sites most often utilized CTOR alone as an image guidance technique, while central nervous system sites most often utilized kV planar imaging alone. Of the patients who required new CT imaging for plan adaptation, 75.0% utilized CTOR images. Some reasons for patients not being able to utilize CTOR imaging for re-planning were the need for 4DCT imaging (which was not available on the CTOR) or the need to re-make immobilization devices (such as thermoplastic masks due to becoming loose). CTOR was shown to be utilized for image guidance in the majority of patients receiving proton therapy at a single-room proton therapy center. CTOR was able to be used for plan adaptation in a large majority of patients, eliminating the need to schedule additional CT simulation appointments for these patients.

Comprehensive Quality of Life Report from a Prospective Clinical Trial of 4DCT-Ventilation Functional Lung Avoidance Radiation Therapy

Functional imaging has been developed that uses 4DCT images and image processing to generate lung ventilation maps (4DCT-ventilation). 4DCT-ventilation functional avoidance uses 4DCT-images to generate plans that avoid functional regions of the lung with the goal of reducing pulmonary toxicity. A 4DCT-ventilation functional avoidance, phase II, multi-center clinical trial was completed, and patient reported outcomes (PRO) measured. PROs are an essential measure of quality-of-life following radiotherapy. The purpose of this work is to quantify PRO changes for lung cancer patients treated with functional avoidance and to compare PROs against clinical pneumonitis. Patients with locally advanced lung cancer receiving curative intent radiotherapy (prescriptions of 45-75 Gy) and chemotherapy were accrued. Each patient had a 4DCT-ventilation image generated using 4DCT data. Favorable arc geometry and optimization techniques were used to generate functional avoidance plans. PRO instruments included the Functional Assessment of Cancer Therapy Lung (FACT-L) questionnaire and the Visual Analog Scale (VAS) administered pre-treatment and 3-, 6-, and 12-months post-treatment to gather data on physical, social, emotional, functional, and pulmonary well-being. The percentage of patients with clinically significant decline was calculated using the FACT-TOI (Trial Outcome Index), FACT-LCS (Lung Cancer Subscale), and VAS instruments. To evaluate the correlation between PROs and clinical toxicity, the percentage of clinically significant FACT-LCS decline was compared (Chi-square test) for patients who did or did not experience grade 2+ pneumonitis. Fifty-nine patients completed baseline PRO surveys. Median age was 65, 83% of patients had non-small-cell lung cancer, with 75% having stage III disease. Clinically significant FACT-TOI decline at, 3, 6, and 12 months was 46.3%, 38.5%, and 26.8%, respectively. The percentage of patients with clinically significant FACT-LCS decline was 33.3%, 33.3%, and 29.3%, at 3 months, 6 months, and 12 months, respectively. The percentage of patients with clinically significant VAS decline at 3, 6, and 12 months was 18.9%, 20.0%, and 18.6%, respectively. Patients who experienced grade 2+ pneumonitis had a greater percentage of clinically significant decline at all time-points with the results reaching significance (p = 0.045) at 6 months. The study presents the first comprehensive evaluation of PROs for patients treated with 4DCT-ventilation functional avoidance. The data show that 20-40% of patients had clinically significant decline and that PROs had a strong correlation with pneumonitis. The PRO data demonstrate that functional avoidance results in low rates of patient reported outcome clinical decline and provide seminal results to be used in phase III studies.

Exploration of AI-Assisted On-Board Perfusion Imaging Technique on Patients Undergoing Thoraxradiotherapy

Pulmonary function of lung cancer patients can significantly change along the radiation therapy (RT) course. Scholars have synthesized lung function maps from computed tomography (CT) images. Still, there is lack of study exploring the feasibility of generating lung perfusion information from cone-beam CT (CBCT). Our study aims to fill up this gap in the body of literatures. One-hundred-and-six pairs of planning CT and CBCT images of thoracic cancer patients from XX hospital were collected retrospectively. All CT images were registered to CBCT with a deformable algorithm, and referred as deformed planning CT (dCT). A 2D convolutional neural network (CNN) model built with cycle generative adversarial network (cycle-GAN) was trained to correct CBCT image artifacts and HU value discrepancies. The corrected CBCT was named enhanced CBCT (eCBCT) in this study. All CBCT, eCBCT and dCT images were then input to a pre-trained CT-to-perfusion deep learning (DL) model to synthesize perfusion images (PI), namely PICBCT, PIeCBCT and PIdCT respectively. For model training, 80 image pairs were assigned to the training set while others to the testing set. dCT and PIdCT were respectively regarded as the ground truth in two consecutive models. One extra validation case acquired with SPECT perfusion scan was collected for model performance test. Quantitative evaluation was done between the ground truth and the synthesized images. Peak signal-to-noise ratio (PSNR) and mean absolute error (MAE) were computed to assess efficacy of CBCT enhancement. Correlation (R) and Dice similarity coefficient (DSC) were used to evaluate voxel-wise and function-wise concurrence in CBCT-derived perfusion mapping. In CBCT enhancement, [eCBCT-dCT] pairs demonstrated higher agreement than [CBCT-dCT] pairs. Quantitatively, PSNR and MAE were improved from 21.10±1.60 to 24.08±1.76, and 68.99±13.51 to 47.06±11.31 (p<0.01), respectively. For perfusion translation, higher correlation was demonstrated in both voxel-wise and function-wise evaluation within [PIeCBCT-PIdCT] than [PICBCT-PIdCT]. R correlation increased from 0.84±0.09 to 0.89±0.06 (p<0.01), whereas DSC in high-functional regions increased from 0.77±0.06 to 0.82±0.05 (p<0.01). In the validation case, strong correlation was observed between SPECT perfusion scan and PIeCBCT, with R increasing from 0.57 to 0.65 when compared to PICBCT. We proposed a novel DL framework that synthesized perfusion images from fractionally acquired CBCT. Our framework met the clinical needs in providing real-time lung perfusion maps. It can be used to track the continuously changing pulmonary function status throughout the RT course using routinely scanned CBCT, without additional imaging modalities. The framework can facilitate functionally guided adaptive radiation therapy by providing immediate pulmonary function information.

Medializing Calcaneal Osteotomy for progressive collapsing foot deformity alters the three-dimensional subtalar joint alignment

BackgroundA medializing calcaneal osteotomy (MCO) is considered as one of the key inframalleolar osteotomies to correct progressive collapsing foot deformity (PCFD). While many studies were able to determine the post-operative hind- and midfoot alignment, alternations of the subtalar joint alignment remained obscured by superposition on plain radiography. Therefore, we aimed to assess the hind-, midfoot- and subtalar joint alignment pre- compared to post-operatively using 3D weightbearing CT (WBCT) imaging. MethodsSeventeen patients with a mean age of 42 ± 17 years were retrospectively analyzed. Inclusion criteria consisted of PCFD deformity corrected by a medializing calcaneal osteotomy (MCO) as main procedure and imaged by WBCT before and after surgery. Exclusion criteria were patients who had concomitant calcaneal lengthening osteotomies, mid-/hindfoot fusions, hindfoot coalitions, and supramalleolar procedures. Image data were used to generate 3D models and compute the hindfoot (HA), midfoot (MA) - and subtalar joint (STJ) alignment in the coronal, sagittal and axial plane, as well as distance maps. ResultsPre-operative measurements of the HA and MA improved significantly relative to their post-operative equivalents p < 0.05). The post-operative STJ alignment showed significant inversion (2.8° ± 1.7), abduction (1.5° ± 1.8), and dorsiflexion (2.3° ± 1.7) of the talus relative to the calcaneus (p < 0.05) compared to the pre-operative alignment. The displacement between the talus and calcaneus relative to the sinus tarsi increased significantly (0.6 mm±0.5; p < 0.05). ConclusionThis study detected significant changes in the sagittal, coronal, and axial plane alignment of the subtalar joint, which corresponded to a decompression of the sinus tarsi. These findings contribute to our clinical practice by demonstrating the magnitude of alteration in the subtalar joint alignment that can be expected after PCFD correction with MCO as main procedure.

Novel Dose Criteria for Lung Cancer SBRT to Improve Local Control in Patients Treated to 50 Gy/5 Fractions Using Deep Learning Methods and Explainability Techniques

To some radiation oncologists, 50 Gy/5 fractions has been considered controversial, as they feel the nominal BED of 100 Gy might be too low for long-term local control of some lesions. We analyzed a large cohort of these patients using a deep learning model to predict local recurrence (LR) and used explainability techniques to extract new dose features important to the model's prediction. Subsequently, we determined optimal cut-points for the most significant metrics to provide actionable criteria for treatment planning in these patients. A total of 535 SBRT lung cancer patients treated between 2009 and 2017 were retrospectively analyzed using a deep learning approach. All patients had NSCLC and all of them were treated with 50 Gy in 5 fractions (100 Gy BED, α/β = 10). Mean clinical maximum tumor diameter was 2.2 cm. There were 31 LR in the dataset with mean follow-up time of 28 months. Mean age was 75 years. CT images, 3D dose distribution and patient demographic details were used to train a deep learning survival model to predict time to failure and probability of local control. Validation, training, and testing were in accordance with TRIPOD criteria. 80 % of the data were used for 5-fold cross-validation (10 iterations) and 20 % was held for independent testing. The Grad-CAM method was applied to identify regions of the dose distribution that are the most significant to the model's decision-making. Based on the results, appropriate dose metrics were proposed, and optimal cut-points were determined to distinguish between lower and higher LR-risk patients. The model has an acceptable performance (c-index: 0.72, 95% CI: 0.68-0.75); the testing c-index was 0.69. Grad-CAM showed that the model's spatial attention was mostly concentrated in the tumor's "PTV-GTV" region. Statistically significant criteria are in Table 1. A novel deep learning model for prediction of LR, incorporating 3D dose data, CT images and patient demographics, was developed and tested. Grad-CAM demonstrated superior significance of peripheral (PTV-GTV) dose features. Subsequently determined optimal cut-points have significant prognostic power (log rank, p<0.001) and could be used as additional criteria in treatment planning. While these data have repercussions in treatment planning, they do not suggest that a significantly higher BED for the prescription dose is necessary for tumor control in NSCLC. Nevertheless, it might be effective to slightly elevate the prescribed dose, i.e., from 100 Gy BED to 104 Gy BED.

Clinical Outcomes and Setup Reproducibility in Patients Receiving Synchronous Bilateral Breast Radiation in the Prone Position

Standardized radiation therapy (RT) positioning and planning are vital to treat both breasts without excess toxicity in patients with synchronous bilateral breast cancer (SBBC). We studied clinical outcomes and setup reproducibility in patients with SBBC treated in the prone position. This is a retrospective study of patients receiving RT for SBBC from 2012-2022 treated in the prone position. Patient demographics, RT dose and field, toxicity, and outcomes were collected. RT planning used gantry and table shifts to avoid entrance/exit through contralateral breast and organs at risk. RT delivery was standardized, with left breast treated first. After 2014, radiographic films were placed during the first 5 fractions to evaluate field overlap. Consistent overlap required re-simulation or re-planning. Positional shifts during setup were collected for patients receiving bilateral whole breast irradiation (WBI) and bilateral partial breast irradiation (PBI). A total of 45 patients were treated for SBBC. Median age was 67 years. Median follow-up was 28 months. 35 patients were treated with WBI (1 with low axilla on one side) and 5 with PBI. 5 patients were treated with WBI in one breast and PBI in the other. The most common whole breast dose was 40.5Gy, with a simultaneous integrated boost to the lumpectomy cavity to 48Gy. PBI patients were treated with 30Gy (n = 4) or 40.05Gy (n = 1). 8 patients (17.7%) had acute grade 2 and 1 patient (2%) had acute grade 3 dermatitis. This patient was treated WBI bilaterally and was known to have overlap on GaF. Physician and patient accepted higher toxicity for adequate coverage of bilateral upper inner quadrant tumors. Patients with grade > = 2 dermatitis had bilateral WBI (+/- low axilla) except 1. 6 patients had acute dermatitis in the sternal area (5 were WBI and 1 PBI). Only 2 of these patients had overlap on GaF. Of 20 patients with late toxicity follow-up (> 6 months), 5 (25%) had late grade 1-2 dermatitis. 4 (20%) were treated with WBI (+/-low axilla). 1 patient had both a local and distant recurrence. For setup and treatment delivery, couch shifts recorded for patients treated with bilateral WBI (n = 33) and bilateral PBI (n = 5) were minimal. Mean shifts in nearly all dimensions were limited to sub-centimeter or sub-degree changes, with 2D and cone beam CT (CBCT) imaging. Analysis of GaF showed that only 3 patients (10%) had field overlap. A new plan was created for 1 patient to increase the gap between fields, and for the others, patient and physician accepted overlap for adequate coverage. To our knowledge, this is the first study examining clinical outcomes and setup reproducibility in patients treated with SBBC in the prone position. This study indicates the feasibility of prone bilateral breast treatment with minimal shifts and overlap on daily treatments. However, we do see a higher number of acute dermatitis in patients treated with bilateral WBI (vs. PBI) and that overlap was not seen on GaF in all patients who developed sternal dermatitis.

Patient-Specific 3D CT Images Reconstruction from 2D KV Images Via Vision Transformer-Based Deep-Learning

In some proton therapy facilities, patient alignment relies on two 2D orthogonal kV images, taken at fixed, oblique angles, as no 3D on-the-bed-imaging is available. The visibility of the tumor in kV images is limited since the patient's 3D anatomy is projected onto a 2D plane, especially when the tumor is behind a high-density structure such as bone. This can lead to a large patient setup error. A solution to this problem is to reconstruct the 3D CT image from the kV images obtained in the treatment position. An asymmetric autoencoder-like network built with vision-transformer blocks was developed. The data was collected from a head and neck patient: 2 orthogonal kV images (1024X1024 voxels), 1 3D CT with padding (512X512X512) acquired from the in-room CT-on-rails before kVs were taken and 2 digitally-reconstructed-radiograph (DRR) images (512X512) based on the CT. We resampled kV images every 8 voxels and DRR and CT every 4 voxels, thus formed a dataset consisting of 262,144 samples, in which the images had a dimension of 128 for each direction. The value of each voxel in CT was normalized to range 0-1 with a uniform shift of 1000 and a denominator of 4000. For kV and DRR, we ranked all voxels value in an ascending order and normalized the values of the first 80% voxels to range 0-0.8 and the rest to range 0.8-1, thus yielding a quasi-Gaussian distribution, which was favorable by the deep neural networks. We further cropped kV and DRR images with a self-supervised bitmap based on the voxels' gradients. In training, both kV and DRR were utilized, and the encoder was encouraged to learn the same feature maps for kV images and its corresponding DRR images with mean-absolute-error (MAE) as the similarity loss. Then the decoder would reconstruct the 3D CT image from the feature maps of the kV images with the CT-on-rails as ground-truth (gCT) and MAE as the reconstruction loss. In testing, only independent kV images were used. The full-size synthetic CT (sCT) was achieved by concatenating the sCTs generated by the model according to their spatial information. The image quality of the sCT was evaluated using MAE and per-voxel-absolute-CT-number-difference volume histogram (CDVH). The proposed network was implemented with PyTorch deep learning library and both distributed data parallel (DDP) and automatic mixed precision (AMP) were applied to saving memory and accelerating the training speed. We used the AdamW optimizer with β1 = 0.9 and β2 = 0.999 and a cosine annealing learning rate scheduler with an initial learning of 1e-7 and 20 warm-up epochs. The model achieved a MAE of <40HU and the CDVH showed that <5% of the voxels had a per-voxel-absolute-CT-number-difference larger than 185HU. The profile of a typical gCT slice and its corresponding sCT slice exhibited a high agreement, indicating the high similarity between the gCT and sCT. A patient-specific vision-transformer-based network was developed and shown to be accurate and efficient to reconstruct 3D CT images from kV images.

Optimizing the Imaging Scheme of Small-Dose Contrast Agent in 4D-CT Localization Enhancement Scan for Radiotherapy of Liver Cancer

The primary liver cancer radiation therapy 4D-CT-enhanced (CE-4DCT) scan can accurately reflect the state of the liver as it moves with respiration, and enables clear visualization of the tumor, providing a precise range for target area outlining. However, the CE-4DCT scan localization time of primary liver cancer is long, and the fixed delay time of diagnostic CT cannot obtain high-quality 4DCT images. To this end, we designed a randomized controlled study to investigate the ideal localization scan delay time parameters for primary liver cancer CE-4DCT. Patients with primary hepatocellular carcinoma who presented to our department for radiotherapy CT localization from January to December 2022 participated in this study. All subjects were randomly divided into two groups in a 1:1 ratio. CE-4DCT scans were performed using a high concentration of iodomepril (400 mg/ml) contrast agent with a contrast volume equal to the patient's body weight. The scanning parameters were set according to scheme A (flow rate 2.0 ml/s, delay time 30s) and B (2.0 ml/s, 35s) respectively. Before the CE-4D CT scan, CT plain scan was routinely performed, physicians measured the CT value (HU) of abdominal aorta, portal vein, left hepatic vein, middle hepatic vein, right hepatic vein, liver parenchyma and tumor lesion on CT40 time item and plain scan images respectively. The image quality was subjectively and qualitatively assessed. The image enhancement effect is divided into excellent, good, relatively poor and poor. A total of 18 patients were included in this study. Patient characteristics were well-balanced in both groups. There were statistical differences between the two groups (A vs B) in the abdominal aorta (226.66 ± 45.59 vs. 176.66 ± 32.84, p = 0.017), left hepatic vein (142.77 ± 24.41 vs. 164.88 ± 19.02, p = 0.048); in the portal vein, middle hepatic vein, right hepatic vein, liver parenchyma, and tumor lesions CT values were not statistically different. The comparison of all flat-scan images between the two groups was statistically significant. The comparison of all flat-scan images in both groups was statistically significant. The subjective qualitative assessment of the developing effect was excellent in 44.4% (33.3% vs. 11.1%), good in 16.7% (5.6% vs. 11.1%), and poor in 38.9% (16.7% vs. 22.2%) in both groups. image quality was slightly higher in group A than in group B, but did not reach statistical significance (χ2 = 1.088, p = 0.698). In patients with primary hepatocellular carcinoma with portal vein carcinoma thrombosis 83.33% subjectively evaluated poor results. Qualitatively acceptable enhanced images were obtained for primary hepatocellular carcinoma (CE-4DCT) localization scans with a delay time controlled at 30s when using the least amount of contrast agent and the same flow rate.

Assessment of Daily Target Motion to Optimize ITV Margins in Online Adaptive Plan for Liver and Pancreatic SBRT on MR-LINAC

Accurate measurement of daily target motion on online plan adaptation is essential for re-optimizing the internal target volume (ITV) margins. In this study, by analyzing 2D cine MR images acquired in a 1.5T MR-Linac, we assessed daily target motion and its inter-fractional variations, which is then used to customize ITV margins for MR-based online adaptive plan of liver and pancreatic stereotactic body radiotherapy (SBRT). Twenty-threepatients, enrolled in this study, underwent MR-guided online adaptive SBRT on 1.5 T MR-Linac: liver (n = 12 without compression belt (CB), n = 2 with CB) and pancreas (n = 9 with CB). Target motion in a 2D coronal and sagittal MR images (> 200 cine images to cover multiple breathing cycles), as well as a set of 4D-CT images during the simulation, was analyzed to determine patient specific ITV margins for a reference plan. On every fraction of online adaptive SBRT, the 2D MR images were exported to commercially available software along with a planning MR images: a T1-wighted MR image set for pancreas and or a navigated T2-wighted MR image set for liver. Daily target motion was then measured by overlaying the orthogonal cine image set on the T1- or T2-wighted image set. The ITV margin was adjusted when required prior to online plan optimization RESULTS: Thisstudy successfully assessed the daily target motion and inter-fractional variations in 23 liver or pancreatic cancer patients. We found a discrepancy of daily target motion of orthogonal 2D cine MR images from that of 4D-CT, required to adjust the ITV margin for online plan adaptation. In liver SBRT without CB, 15 out of 45 fractions from 12 patients (33.3 %) went through adjusted ITV margin. The tumor motion was large up to a 2.2 cm along the inferior direction in a patient without CB and thus the patient was treated on a conventional Linac with the active breathing control (ABC) unit for breath-hold. On the other hand, no adjustments were necessary for 2 liver patients with CB. In 9 patients with pancreatic cancers, we found the inter-fractional ITV variations in 3 out of 45 fractions (6.7 %) adjusted the ITV margin. The daily target motion in the pancreatic patients with CB was more stable and relatively smaller than that of the liver patients. Theorthogonal 2D cine MR images are evidently essential for assessing daily target motion to evaluate the inter-fractional variation and achieve daily ITV-customized liver and pancreatic SBRT in the 1.5 T MR-Linac. The customized ITV approach will improve the treatment outcome of the MR-guided online adaptive therapy.

4D Monte Carlo dose reconstructions using surface motion measurements

PurposeTo investigate the feasibility of a 4D Monte Carlo based dose reconstruction method to study the dosimetric impact of respiratory motion using surface motion measurements for patients undergoing VMAT treatments for Non-Small Cell Lung Cancer. MethodsThe 4Ddefdosxyznrc/EGSnrc algorithm was used to reconstruct VMAT doses delivered to the patients using machine log files and respiratory traces measured with the RADPOS 4D dosimetry system. The RADPOS sensor was adhered to the patient’s abdomen prior to each treatment fraction and its position was used as a surrogate for tumour motion. Treatment log files were synchronized with the patient respiratory traces. Patient specific respiratory models were generated from deformable registration of the inhale and exhale 4DCT images and the respiratory traces.The reconstructed doses were compared to planned doses calculated with DOSXYZnrc/EGSnrc on the average-intensity and the exhale phase CT images. ResultsRespiratory motion measurements and log files were acquired for 2 patients over 5 treatment fractions each. The motion was predominantly along the anterior/posterior direction (A/P). The average respiratory amplitudes were 8.7 ± 2.7 mm and 10.0 ± 1.2 mm for Patient 1 and 2, respectively. Both patients displayed inter- and intra-fractional variations in the baseline position. Small inter-fractional differences were observed in the reconstructed doses for each patient. Differences between the reconstructed and planned doses were attributed to differences in organ volumes. ConclusionThe 4D reconstruction method was successfully implemented for the two patients studied. Small differences between the planned and reconstructed doses were observed due to the small tumour motion of these patients.

Investigation of uncertainty in internal target volume definition for lung stereotactic body radiotherapy

This study evaluated the validity of internal target volumes (ITVs) defined by three- (3DCT) and four-dimensional computed tomography (4DCT), and subsequently compared them with actual movements during treatment. Five patients with upper lobe lung tumors were treated with stereotactic body radiotherapy (SBRT) at 48 Gy in four fractions. Planning 3DCT images were acquired with peak-exhale and peak-inhale breath-holds, and 4DCT images were acquired in the cine mode under free breathing. Cine images were acquired using an electronic portal imaging device during irradiation. Tumor coverage was evaluated based on the manner in which the peak-to-peak breathing amplitude on the planning CT covered the range of tumor motion (± 3 SD) during irradiation in the left–right, anteroposterior, and cranio-caudal (CC) directions. The mean tumor coverage of the 4DCT-based ITV was better than that of the 3DCT-based ITV in the CC direction. The internal margin should be considered when setting the irradiation field for 4DCT. The proposed 4DCT-based ITV can be used as an efficient approach in free-breathing SBRT for upper-lobe tumors of the lung because its coverage is superior to that of 3DCT.

Microscopic failure characteristics and critical length of short glass fiber reinforced composites

Short glass fiber reinforced composites (SGFRC) are frequently used to manufacture parts with complex surfaces by injection or compression molding. It was found that the length of fibers is often reduced during the manufacturing processes. To understand the causes of the change in fiber length, ex-situ 3D CT images are taken from tensile experiments to extract the information of the microstructure of the SGFRC, including fiber length, orientation and breakage. In the calculation of the critical length, the Kelly model is modified to include the effect of the stresses at fiber ends and the mechanical effect of internal pore defects on the composites. It is found that the microscopic interaction of short fibers breaks longer fibers, until they are shorter than the critical length. Through comparisons, it is found that the modified Kelly model is more accurate in predicting critical length of short fibers. In this study, micro-failure modes, such as fiber pulling out, interface debonding, matrix damage and pore destruction, are all characterized by micro-CT and scanning electronic microscope to analyze the complex failure mechanisms. The micro failure modes are mainly attributed to the micro geometric features and mass fraction of short fibers.

Deformable lung 4DCT image registration via landmark-driven cycle network.

An automated, accurate, and efficient lung four-dimensional computed tomography (4DCT) image registration method is clinically important to quantify respiratory motion for optimal motion management. The purpose of this work is to develop a weakly supervised deep learning method for 4DCT lung deformable image registration (DIR). The landmark-driven cycle network is proposed as a deep learning platform that performs DIR of individual phase datasets in a simulation 4DCT. This proposed network comprises a generator and a discriminator. The generator accepts moving and target CTs as input and outputs the deformation vector fields (DVFs) to match the two CTs. It is optimized during both forward and backward paths to enhance the bi-directionality of DVF generation. Further, the landmarks are used to weakly supervise the generator network. Landmark-driven loss is used to guide the generator's training. The discriminator then judges the realism of the deformed CT to provide extra DVF regularization. We performed four-fold cross-validation on 10 4DCT datasets from the public DIR-Lab dataset and a hold-out test on our clinic dataset, which included 50 4DCT datasets. The DIR-Lab dataset was used to evaluate the performance of the proposed method against other methods in the literature by calculating the DIR-Lab Target Registration Error (TRE). The proposed method outperformed other deep learning-based methods on the DIR-Lab datasets in terms of TRE. Bi-directional and landmark-driven loss were shown to be effective for obtaining high registration accuracy. The mean and standard deviation of TRE for the DIR-Lab datasets was 1.20±0.72mm and the mean absolute error (MAE) and structural similarity index (SSIM) for our datasets were 32.1±11.6 HU and 0.979±0.011, respectively. The landmark-driven cycle network has been validated and tested for automatic deformable image registration of patients' lung 4DCTs with results comparable to or better than competing methods.