Digitally Reconstructed Radiographs Research Articles

Multisequence MR-generated sCT is promising for HNC MR-only RT: A comprehensive evaluation of previously developed sCT generation networks.

To verify the feasibility of our in-house developed multisequence magnetic resonance (MR)-generated synthetic computed tomography (sCT) for accurate dose calculation and fractional positioning for head and neck MR-only radiation therapy (RT). Forty-five patients with nasopharyngeal carcinoma were retrospectively studied. By applying our previously in-house developed network, a patient's sCT can rapidly be generated with respect to feeding the sole T1 image, T1C image, T1DixonC image, T2 image, and their combination (five pipelines in total). A k(5)-fold strategy was implemented during model establishment. Dose recalculation was performed for each pipeline generation to attain a dosimetric feasibility evaluation. Fractional positioning evaluation was performed by calculating the digitally reconstructed radiograph (DRR) of the sCT and planning CT and their offset to the portal image. The dose mean absolute error values were (0.47±0.16)%, (0.48±0.15)% (p<0.05), (0.50±0.16)% (p<0.05), (0.50±0.15)% (p<0.05), and (0.45±0.16)% (p<0.05) for the T1, T1C, T1Dixon C, T2, and 4-channel generated sCT to the prescription dose, respectively. The 4-channel-generated sCT outperforms any other single-sequence pipeline. Among the single-sequence MR imaging-generated sCTs, the T1-generated sCT shows the most accurate HU image quality and provides a reliable dose result. Quantified positioning errors with calculation of the difference to the planning CT offsets are (-0.26±0.50) mm, (-0.58±0.52) mm (p<0.05), (-0.27±0.57) mm (p>0.05), (-0.31±0.44) mm (p>0.05), and (-0.19±0.37) mm (p>0.05) at LNG and (0.34±0.53) mm, (0.48±0.56) mm (p>0.05), (0.55±0.56) mm (p>0.05), (0.37±0.61) mm (p>0.05), and (0.24±0.43) mm (p>0.05) at LAT of the anterior-posterior direction for the five pipelines. Multisequence MR-generated sCT allows for accurate dose calculation and fractional positioning for head and neck MR-only RT.

The Relationship between Lung and Heart Two-Dimensional Parameters and Three-Dimensional Dose-Volume Data in Adjuvant Radiotherapy for Breast Cancer.

Background: Two-dimensional (2D) radiographic parameters have been used to estimate the amount of heart and lung irradiated for minimizing heart and lung complications in breast cancer patients. The aim of this study was to investigate the correlation between traditionally used 2D radiographic and dose-volume parameters during adjuvant radiotherapy of breast cancer. Methods: In this cross-sectional study, we analyzed 121 female patients treated with breast-conserving surgery (BCS) or modified radical mastectomy (MRM) and 3D conformal radiotherapy (3DCRT) using two-field radiotherapy (2FRT) or three-field radiotherapy (3FRT) technique. All patients underwent computed tomography (CT)-planning. Two-D parameters, including central lung distance (CLD), maximum lung depth (MLD), maximum heart length (MHL), maximum heart distance (MHD), and chest wall separation (CWS), were measured using digitally reconstructed radiographs (DRR) and CT images. DVHs for lung, heart, and target were created. The Pearson correlation test was used to evaluate the correlation between 2D radiographic and dose-volume parameters. Results: There was a correlation between CLD and ipsilateral lung V5-20Gy and Dmean and between MLD and ipsilateral lung V5-20Gy. In 2FRT, only moderate correlation between CLD and ipsilateral lung V20Gy (r = 0.453, P = 0.003) and between MLD and ipsilateral lung V20Gy (r = 0.593, P <0.001) were observed. Poor correlation of MHL and heart V25Gy (r = 0.409, P = 0.007) was seen only in 3FRT. There was a correlation between MHD and heart dose-volume data, with a strong correlation between MHD and heart V5-25Gy and Dmean (r = 0.875-0.934, P<0.001) in the 2FRT group. No correlation between CWS and breast Dmax was found. Conclusion: There was a correlation between 2D parameters (i.e., CLD, MLD, and MHD) and the heart and lung dose-volume parameters during adjuvant breast radiotherapy. Although CLD was correlated to ipsilateral lung V5-20Gy and Dmean, the correlation between CLD and ipsilateral lung V20Gy was greater than other dose-volume parameters. MHD provided a close estimation of heart dose-volume parameters.

Offline generator for digitally reconstructed radiographs of a commercial stereoscopic radiotherapy image-guidance system.

PurposeImage‐guided radiotherapy (IGRT) research sometimes involves simulated changes to patient positioning using retrospectively collected clinical data. For example, researchers may simulate patient misalignments to develop error detection algorithms or positioning optimization algorithms. The Brainlab ExacTrac system can be used to retrospectively “replay” simulated alignment scenarios but does not allow export of digitally reconstructed radiographs (DRRs) with simulated positioning variations for further analysis. Here we describe methods to overcome this limitation and replicate ExacTrac system DRRs by using projective geometry parameters contained in the ExacTrac configuration files saved for every imaged subject.MethodsTwo ExacTrac DRR generators were implemented, one with custom MATLAB software based on first principles, and the other using libraries from the Insight Segmentation and Registration Toolkit (ITK). A description of perspective projections for DRR rendering applications is included, with emphasis on linear operators in real projective space P3. We provide a general methodology for the extraction of relevant geometric values needed to replicate ExacTrac DRRs. Our generators were tested on phantom and patient images, both acquired in a known treatment position. We demonstrate the validity of our methods by comparing our generated DRRs to reference DRRs produced by the ExacTrac system during a treatment workflow using a manual landmark analysis as well as rigid registration with the elastix software package.ResultsManual landmarks selected between the corresponding DRR generators across patient and phantom images have an average displacement of 1.15 mm. For elastix image registrations, we found that absolute value vertical and horizontal translations were 0.18 and 0.35 mm on average, respectively. Rigid rotations were within 0.002 degrees.ConclusionCustom and ITK‐based algorithms successfully reproduce ExacTrac DRRs and have the distinctive advantage of incorporating any desired 6D couch position. An open‐source repository is provided separately for users to implement in IGRT patient positioning research.

Development of a deep learning‐based patient‐specific target contour prediction model for markerless tumor positioning

For pancreatic cancer patients, image guided radiation therapy and real-time tumor tracking (RTTT) techniques can deliver radiation to the target accurately. Currently, for the radiation therapy machine with kV X-ray imaging systems, internal markers must be implemented to facilitate tumor tracking. The purpose of this study was to develop a markerless deep learning-based pancreatic tumor positioning procedure for real-time tumor tracking with a kV X-ray imaging system. Fourteen pancreatic cancer patients treated with intensity-modulated radiation therapy from six fixed gantry angles with a gimbal-head radiotherapy system were included in this study. For a gimbal-head radiotherapy system, the three-dimensional (3D) intrafraction target position can be determined using an orthogonal kV X-ray imaging system. All patients underwent four-dimensional computed tomography (4DCT) simulations for treatment planning, which were divided into 10 respiratory phases. After a patient's 4DCT was acquired, for each X-ray tube angle, 10 digitally reconstructed radiograph (DRR) images were obtained. Then, a data augmentation procedure was conducted. The data augmentation procedure first rotated the CT volume around the superior-inferior and anterior-posterior directions from -3° to 3° in 1.5° intervals. Then, the Super-SloMo model was adapted to interpolate 10 frames between respiratory phases. In total, the data augmentation procedure expanded the data scale 250-fold. In this study, for each patient, 12 datasets containing the DRR images from each specific X-ray tube angle based on the radiation therapy plan were obtained. The augmented dataset was randomly divided into training and testing datasets. The training dataset contained 2000 DRR images with clinical target volume (CTV) contours labeled for fine-tuning the pre-trained target contour prediction model. After the fine-tuning, the patient and X-ray tube angle-specific CTV contour prediction model was acquired. The testing dataset contained the remaining 500 images to evaluate the performance of the CTV contour prediction model. The dice similarity coefficient (DSC) between the area enclosed by the CTV contour and predicted contour was calculated to evaluate the model's contour prediction performance. The 3D position of the CTV was calculated based on the centroid of the contour in the orthogonal DRR images, and the 3D error of the prediction position was calculated to evaluate the CTV positioning performance. For each patient, the DSC results from 12 X-ray tube angles and 3D error from 6 gantry angles were calculated, representing the novelty of this study. The mean and standard deviation (SD) of all patients' DSCs were 0.98 and 0.015, respectively. The mean and SD of the 3D error were 0.29mm and 0.14mm, respectively. The global maximum 3D error was 1.66mm, and the global minimum DSC was 0.81. The mean calculation time for CTV contour prediction was 55ms per image. This fulfills the requirement of RTTT. Regarding the positioning accuracy and calculation efficiency, the presented procedure can provide a solution for markerless real-time tumor tracking for pancreatic cancer patients.

Reproducibility of deep inspiration breath-hold technique for left-side breast cancer with respiratory monitoring device, Abches.

PurposeThis study aimed to evaluate the reproducibility of deep inspiration breath‐hold (DIBH) using a respiratory control device, Abches, in patients with left‐sided breast cancer.Material and methodsAbches comprises a main body, an indicator panel, and two fulcrums, one each on the chest and abdomen. Forty left side breast cancer patients treated with DIBH using abches were enrolled in this study. For all patients, CT images were taken three times to confirm the target position inside the body and to check the breath‐hold reproducibility. Three anatomical points on the nipple, sternum, and heart were selected as measurement points on CT images. After measuring the coordinates, breath‐hold reproducibility was defined as the mean of the absolute difference in the coordinates between the three CT images. The maximum differences were also investigated. In addition, the dice similarity coefficient (DSC) was calculated to examine the displacement of the heart volume in detail. Moreover, digitally reconstructed radiographs (DRRs) and linac graphs (LGs) were used to measure the positional accuracy of the chest and heart.ResultsThe reproducibility in all patients was within 0.75 mm for the nipple, 0.78 mm for the sternum, and 2.18 mm for the heart in each direction. Similarly, the maximum displacements for all patients were within 1.90 mm, 1.69 mm, and 4.75 mm, respectively, in each direction. For heart volume, the average DSC for all cases was 0.93 ± 0.01. The differences between the DRR and LG images were 1.70 ± 1.10 mm and 2.10 ± 1.60 mm, for the chest and heart, respectively.ConclusionOur study showed that DIBH using Abches can be performed with good target reproducibility of less than 3 mm with proper breath‐hold practice, whereas the heartbeat‐derived reproducibility of the cardiac position is poor and needs to be monitored carefully during treatment simulation

Bone suppression on pediatric chest radiographs via a deep learning-based cascade model

Background and objectiveBone suppression images (BSIs) of chest radiographs (CXRs) have been proven to improve diagnosis of pulmonary diseases. To acquire BSIs, dual-energy subtraction (DES) or a deep-learning-based model trained with DES-based BSIs have been used. However, neither technique could be applied to pediatric patients owing to the harmful effects of DES. In this study, we developed a novel method for bone suppression in pediatric CXRs. MethodsFirst, a model using digitally reconstructed radiographs (DRRs) of adults, which were used to generate pseudo-CXRs from computed tomography images, was developed by training a 2-channel contrastive-unpaired-image-translation network. Second, this model was applied to 129 pediatric DRRs to generate the paired training data of pseudo-pediatric CXRs. Finally, by training a U-Net with these paired data, a bone suppression model for pediatric CXRs was developed. ResultsThe evaluation metrics were peak signal to noise ratio, root mean absolute error and structural similarity index measure at soft-tissue and bone region of the lung. In addition, an expert radiologist scored the effectiveness of BSIs on a scale of 1–5. The obtained result of 3.31 ± 0.48 indicates that the BSIs show homogeneous bone removal despite subtle residual bone shadow. ConclusionOur method shows that the pixel intensity at soft-tissue regions was preserved, and bones were well subtracted; this can be useful for detecting early pulmonary disease in pediatric CXRs.

Differences in image density adjustment parameters on the image matching accuracy of a floor-mounted kV X-ray image-guided radiation therapy system.

This study aimed to investigate the effect of two different image density adjustment parameters on the results of image matching at six degrees of freedom using radiographic images generated by the ExacTrac X‐ray system in brain stereotactic radiosurgery (SRS). This study comprised 32 patients who underwent brain SRS at our hospital from January 2020 to December 2020. In this study, (1) the default parameter (an image density parameter between “tissue” and “bone”) was an image density parameter for digitally reconstructed radiograph (DRR) generation used at many facilities, and (2) the bone parameter was the steepest contrast parameter used at our hospital. Of the 32 patients, 24 (75%) had a couch angle of 0.5 mm or more in the translational direction or 0.5° or more in the rotational direction, and 10 (31%) had a couch angle of 1.0 mm or more in the translational direction or 1.0° or more in the rotational direction. Among the 131 cases of all couch angles, 46 (35%) cases had a translational direction of 0.5 mm or more or a rotational direction of 0.5° or more, and 15 (11%) had a translational direction of 1.0 mm or more or a rotational direction of 1.0° or more. The results of this study indicate the usefulness of using appropriate DRR parameters for each case, rather than using the default settings. The use of appropriate DRR parameters can lead to accurate position matching results, leading to fewer image‐guided radiation therapy shots and a lower imaging dose.

Decompose kV projection using neural network for improved motion tracking in paraspinal SBRT.

On-treatment kV images have been used in tracking patient motion. One challenge of markerless motion tracking in paraspinal SBRT is the reduced contrast when the X-ray beam needs to pass through a large portion of the patient's body, for example, from the lateral direction. Besides, due to the spine's overlapping with the surrounding moving organs in the X-ray images, auto-registration could lead to potential errors. This work aims to automatically extract the spine component from the conventional 2D X-ray images, to achieve more robust and more accurate motion management. A ResNet generative adversarial network (ResNetGAN) consisting of one generator and one discriminator was developed to learn the mapping between 2D kV image and the reference spine digitally reconstructed radiograph (DRR). A tailored multi-channel multi-domain loss function was used to improve the quality of the decomposed spine image. The trained model took a 2D kV image as input and learned to generate the spine component of the X-ray image. The training dataset included 1347 2D kV thoracic and lumbar region X-ray images from 20 randomly selected patients, and the corresponding matched reference spine DRR. Another 226 2D kV images from the remaining four patients were used for evaluation. The resulted decomposed spine images and the original X-ray images were registered to the reference spine DRRs, to compare the spine tracking accuracy. The decomposed spine image had the mean peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) of 60.08 and 0.99, respectively, indicating the model retained and enhanced the spine structure information in the original 2D X-ray image. The decomposed spine image matching with the reference spine DRR had submillimeter accuracy (in mm) with a mean error of 0.13, 0.12, and a maximum of 0.58, 0.49 in the - and -directions (in the imager coordinates), respectively. The accuracy improvement is robust in all lateral and anteroposterior X-ray beam angles. We developed a deep learning-based approach to remove soft tissues in the kV image, leading to more accurate spine tracking in paraspinal SBRT.

Association of a Deep Learning Estimation of Chest Imaging Age With Survival in Patients With Non-Small Cell Lung Cancers Undergoing Radiation

Recent work has challenged emphasis of chronologic age in cancer treatment decision-making. For patients with non-small cell lung cancer (NSCLC), cardiopulmonary risk histories widely vary, possibly impacting "biologic age." We aimed to apply an externally-derived deep learning model to estimate chest imaging age (ChIA) from digitally reconstructed radiographs (DRRs) created from simulation computed tomography (CT) scans of patients with locally advanced NSCLC (LA-NSCLC). We hypothesized ChIA could serve as a prognostic imaging biomarker in this population.We previously pretrained a Resnet34 convolutional neural network (CNN) to predict chronologic age using normal chest radiographs (CXRs) from public cohorts (CheXpert, NIH, and PadCHEST, n = 24,934). We fine-tuned the CNN to predict ChIA based on time to death using CXRs from 13,657 (25%) patients of the Prostate, Lung, Colorectal and Ovarian Cancer Trial, testing in the remaining 75% (n = 40,967) and the National Lung Screening Trial (n = 5,414). We applied the model to simulation CT DRRs in an institutional cohort of patients with LA-NSCLC receiving lung radiation (Lung-RT, n = 847). Recursive partitioning analysis (RPA) identified a cut point associated with overall survival (OS) in Lung-RT, above which patients were classified ChIA-high. We evaluated association between ChIA and OS using Kaplan-Meier calculations and Cox regression analyses and then validated in the multi-institutional RTOG 0617 trial (n = 460).In Lung-RT, median age was 67 (Range 21-90) years and median ChIA was 67 (Range 57-80) years. In RTOG 0617, median age was 64 (37-83) years, and median ChIA was 66 (52-77) years. Age/ChIA correlation coefficients were 0.29 (95% CI 0.23-0.35, P < 0.001) and 0.32 (95% CI 0.23-0.40, P < 0.001), respectively. Table 1 shows OS after median follow-up in Lung-RT of 56 (95% CI 53-58) months, and in RTOG 0617 of 36 (95% CI 32-37) months. In Lung-RT, adjusting for age, race, and sex, ChIA-high patients had HR 1.22 (95% CI 1.00-1.49, P = 0.049) for OS. In RTOG 0617, adjusting for trial arm, age, gender, race, ethnicity, performance status, histology, and stage, ChIA-high patients had HR 1.41 (95% CI 1.02-1.91, P = 0.037) for OS.Our work validates use of a deep learning CXR-based model to predict ChIA using DRRs created from routine simulation CTs of patients with LA-NSCLC. In an institutional and multi-institutional trial cohort, ChIA-high patients had worse survival than ChIA-low patients, suggesting potential use of ChIA as a prognostic imaging biomarker enabling de-emphasis of chronologic age in treatment decision-making.

Siamese-Based Deep Learning for Markerless Lung Tumor Tracking During Stereotactic Radiotherapy

<h3>Purpose/Objective(s)</h3> During stereotactic lung radiotherapy we want to monitor the target position to ensure it remains inside the planning target volume (PTV). This ensures accurate treatment delivery and may allow for margin reduction. Many LINAC systems have a kilo-voltage (kV) imaging system mounted orthogonally to the MV beam. This can acquire images during volumetric modulated arc therapy delivery. The 2D-kV images can be used to identify the tumor position, e.g., using template matching and triangulation (TMT). However, small lung targets are difficult to track, due to e.g., low tumor contrast, occlusion (e.g., by spine), or noisy images. We investigated deep learning (DL) as an alternative method for lung tumor tracking. <h3>Materials/Methods</h3> Clinical data from 6 patients (P1-P6) with small lung tumors (0.5-3 cm diameter) and a 3D-printed anthropomorphic phantom (P0), including a low-density lung tumor, was used. For each patient, data included 4D-CT scans, fluoroscopic kV images and breathing-related surface motion (Real-Time Position Management, RPM). For the phantom, a 3D-CT scan and kV images taken with a static and moving couch (regular and irregular motion) were available. Due to lack of ground truth tumor locations for the kV images, a patient-specific DL model is trained on Digitally Reconstructed Radiographs (DRRs), generated from the planning CT scans. To enhance the tumor's visual diversity in DRRs, we augmented the training data with locally deformed CT scans. A Siamese-based tracking by similarity method is hypothesized to be better suited for the tumor tracking task than other DL methods, due to its more localized search space and the use of previous frames (temporal information). <h3>Results</h3> Table 1 shows the superior-inferior (SI) direction results obtained by the Siamese model and TMT, on all kV images. For the patient data, the <b>Correlation coefficient</b> between estimated tumor trajectory and RPM was computed, due to lack of ground truth locations. For the phantom, the <b>Mean Absolute Distance</b> (MAD) between detected and ground truth center coordinates was computed. For all 6 patients and the phantom, the Siamese model outperformed TMT. <h3>Conclusion</h3> A novel approach for real-time markerless lung tumor tracking is presented, based on deep Siamese networks. The model compares favorably to the template matching/triangulation-based technique and merits further investigation. The current analysis presents only SI tracking. Reliable 3D tracking, with sub-mm accuracy, is the ultimate goal. Further work is needed to achieve this.

Transfer learning from an artificial radiograph-landmark dataset for registration of the anatomic skull model to dual fluoroscopic X-ray images

Registration of 3D anatomic structures to their 2D dual fluoroscopic X-ray images is a widely used motion tracking technique. However, deep learning implementation is often impeded by a paucity of medical images and ground truths. In this study, we proposed a transfer learning strategy for 3D-to-2D registration using deep neural networks trained from an artificial dataset. Digitally reconstructed radiographs (DRRs) and radiographic skull landmarks were automatically created from craniocervical CT data of a female subject. They were used to train a residual network (ResNet) for landmark detection and a cycle generative adversarial network (GAN) to eliminate the style difference between DRRs and actual X-rays. Landmarks on the X-rays experiencing GAN style translation were detected by the ResNet, and were used in triangulation optimization for 3D-to-2D registration of the skull in actual dual-fluoroscope images (with a non-orthogonal setup, point X-ray sources, image distortions, and partially captured skull regions). The registration accuracy was evaluated in multiple scenarios of craniocervical motions. In walking, learning-based registration for the skull had angular/position errors of 3.9 ± 2.1°/4.6 ± 2.2 mm. However, the accuracy was lower during functional neck activity, due to overly small skull regions imaged on the dual fluoroscopic images at end-range positions. The methodology to strategically augment artificial training data can tackle the complicated skull registration scenario, and has potentials to extend to widespread registration scenarios.

Synthetic digital reconstructed radiographs for MR-only robotic stereotactic radiation therapy: A proof of concept

PurposeTo create synthetic CTs and digital reconstructed radiographs (DRRs) from MR images that allow for fiducial visualization and accurate dose calculation for MR-only radiosurgery. MethodsWe developed a machine learning model to create synthetic CTs from pelvic MRs for prostate treatments. This model has been previously proven to generate synthetic CTs with accuracy on par or better than alternate methods, such as atlas-based registration. Our dataset consisted of 11 paired CT and conventional MR (T2) images used for previous CyberKnife (Accuray, Inc) radiotherapy treatments. The MR images were pre-processed to mimic the appearance of fiducial-enhancing images. Two models were trained for each parameter case, using a sub-set of the available image pairs, with the remaining images set aside for testing and validation of the model to identify the optimal patch size and number of image pairs used for training. Four models were then trained using the identified parameters and used to generate synthetic CTs, which in turn were used to generate DRRs at angles 45° and 315°, as would be used for a CyberKnife treatment. The synthetic CTs and DRRs were compared visually and using the mean squared error and peak signal-to-noise ratio against the ground-truth images to evaluate their similarity. ResultsThe synthetic CTs, as well as the DRRs generated from them, gave similar visualization of the fiducial markers in the prostate as the true counterparts. There was no significant difference found for the fiducial localization for the CTs and DRRs. Across the 8 DRRs analyzed, the mean MSE between the normalized true and synthetic DRRs was 0.66 ± 0.42% and the mean PSNR for this region was 22.9 ± 3.7 dB. For the full CTs, the mean MAE was 72.9 ± 88.1 HU and the mean PSNR was 31.2 ± 2.2 dB. ConclusionsOur machine learning-based method provides a proof of concept of a way to generate synthetic CTs and DRRs for accurate dose calculation and fiducial localization for use in radiation treatment of the prostate.

An Ultra-Fast Digitally Reconstructed Radiograph (DRR) Implementation of the Siddon-Jacobs Algorithm using Parallel Computing: Runtime Improvement of an Intensity-Based 2D/3D Registration

Abstract A key component of an intensity-based 2D/3D registration is the digitally reconstructed radiograph (DRR) module, which creates 2D projections from pre-operative 3D data, e.g., CT and MRI scans. On average, an intensity-based 2D/3D registration requires ten iterations and the rendering of twelve DRR images per iteration. In a typical DRR implementation, the rendering time is about two seconds, and the registration runtime is four minutes. We present an implementation of the Siddon-Jacobs algorithm that uses a novel pixel-step approach to determine the pixel location of the rendering plane. In addition, we calculate the intensity of each pixel in the rendering plane using a parallel computing approach. The DRR rendering time is reduced to 10ms on average so that the registration runtime can be achieved in an average of 4.8 seconds.

CrossModalNet: exploiting quality preoperative images for multimodal image registration

A long-standing problem in image-guided radiotherapy is that inferior intraoperative images present a difficult problem for automatic registration algorithms. Particularly for digital radiography (DR) and digitally reconstructed radiograph (DRR), the blurred, low-contrast, and noisy DR makes the multimodal registration of DR–DRR challenging. Therefore, we propose a novel CNN-based method called CrossModalNet to exploit the quality preoperative modality (DRR) for handling the limitations of intraoperative images (DR), thereby improving the registration accuracy. The method consists of two parts: DR–DRR contour predictions and contour-based rigid registration. We have designed the CrossModal Attention Module and CrossModal Refine Module to fully exploit the multiscale crossmodal features and implement the crossmodal interactions during the feature encoding and decoding stages. Then, the predicted anatomical contours of DR–DRR are registered by the classic mutual information method. We collected 2486 patient scans to train CrossModalNet and 170 scans to test its performance. The results show that it outperforms the classic and state-of-the-art methods with 95th percentile Hausdorff distance of 5.82 pixels and registration accuracy of 81.2%. The code is available at https://github.com/lc82111/crossModalNet.

Conception of a C-Arm Parametrization Device Evaluated by an Inverse Registration Approach: An application for 2D/3D Registration in Minimally Invasive Surgery

Abstract Camera resectioning is the process of finding the intrinsic and extrinsic parameters of the pinhole camera model. A mobile X-ray device (C-arm) can be represented with the pinhole camera model. The C-arm parameters can be determined with a parametrization device and the direct linear transformation (DLT) algorithm. We present the conception process of a parametrization device and further optimizations of the DLT algorithm. The evaluation of the parametrization quality is carried out with an inverse registration approach on a testing device. The seven landmarks in the testing device are compared on an X-ray image and a digitally reconstructed radiograph (DRR). The average distance of the corresponding landmarks between imaging modalities is 2.5 pixels. This deviation indicates an adequate design of the parametrization device for a 2D/3D registration in minimally invasive surgery.

2D\u20133D reconstruction of distal forearm bone from actual X-ray images of the wrist using convolutional neural networks

The purpose of the study was to develop a deep learning network for estimating and constructing highly accurate 3D bone models directly from actual X-ray images and to verify its accuracy. The data used were 173 computed tomography (CT) images and 105 actual X-ray images of a healthy wrist joint. To compensate for the small size of the dataset, digitally reconstructed radiography (DRR) images generated from CT were used as training data instead of actual X-ray images. The DRR-like images were generated from actual X-ray images in the test and adapted to the network, and high-accuracy estimation of a 3D bone model from a small data set was possible. The 3D shape of the radius and ulna were estimated from actual X-ray images with accuracies of 1.05 ± 0.36 and 1.45 ± 0.41 mm, respectively.

Clinical implementation of automated treatment planning for whole-brain radiotherapy.

The purpose of the study was to develop and clinically deploy an automated, deep learning‐based approach to treatment planning for whole‐brain radiotherapy (WBRT). We collected CT images and radiotherapy treatment plans to automate a beam aperture definition from 520 patients who received WBRT. These patients were split into training (n = 312), cross‐validation (n = 104), and test (n = 104) sets which were used to train and evaluate a deep learning model. The DeepLabV3+ architecture was trained to automatically define the beam apertures on lateral‐opposed fields using digitally reconstructed radiographs (DRRs).For the beam aperture evaluation, 1st quantitative analysis was completed using a test set before clinical deployment and 2nd quantitative analysis was conducted 90 days after clinical deployment. The mean surface distance and the Hausdorff distances were compared in the anterior‐inferior edge between the clinically used and the predicted fields. Clinically used plans and deep‐learning generated plans were evaluated by various dose–volume histogram metrics of brain, cribriform plate, and lens.The 1st quantitative analysis showed that the average mean surface distance and Hausdorff distance were 7.1 mm (±3.8 mm) and 11.2 mm (±5.2 mm), respectively, in the anterior–inferior edge of the field. The retrospective dosimetric comparison showed that brain dose coverage (D99%, D95%, D1%) of the automatically generated plans was 29.7, 30.3, and 32.5 Gy, respectively, and the average dose of both lenses was up to 19.0% lower when compared to the clinically used plans. Following the clinical deployment, the 2nd quantitative analysis showed that the average mean surface distance and Hausdorff distance between the predicted and clinically used fields were 2.6 mm (±3.2 mm) and 4.5 mm (±5.6 mm), respectively.In conclusion, the automated patient‐specific treatment planning solution for WBRT was implemented in our clinic. The predicted fields appeared consistent with clinically used fields and the predicted plans were dosimetrically comparable.

Evaluation of a Weightbearing CT Artificial Intelligence-Based Automatic Measurement for the M1-M2 Intermetatarsal Angle in Hallux Valgus.

Weightbearing cone beam computed tomography (WBCT) has been gaining traction as a useful imaging modality in the diagnosis and follow-up of foot and ankle musculoskeletal pathologies due to the ability to perform quick, low-dose, 3-dimensional (3D) scans. However, the resulting wealth of 3D data renders daily clinical use time-consuming. The aim of this study was to evaluate a new artificial intelligence (AI)-based automatic measurement for the M1-M2 intermetatarsal angle (IMA) in hallux valgus (HV). We hypothesized that automatic and manual measurements would have a strong correlation, and that the AI software would yield better reproducibility and would be faster compared with manual measurements. This was a multicenter retrospective comparative case-control study in which a total of 128 feet were included from 93 patients who underwent WBCT scans as part of their routine follow-up: 59 feet with symptomatic HV and 69 controls. The IMA was measured automatically using the AI software and manually on digitally reconstructed radiographs (DRRs). The AI software produced both an automatic 2D (auto 2D) and 3D (auto 3D) measurement. There were strong intermethod correlations between the DRR IMA and the auto 2D (HV, r = 0.61; control, r = 0.60; all P < .0001) and auto 3D (HV, r = 0.63; control, r = 0.52; all P < .0001) measurements, respectively. The intrasoftware reproducibility was very close to 100%. Measurements took 23.6 ± 2.31 seconds and 14.5 ± 1.18 seconds, respectively, when taken manually on DRRs and automatically. Controls demonstrated a mean DRR IMA of 8.6 (95% CI, 8.1-9.1), mean auto 2D of 11.2 (95% CI, 10.7-11.7), and mean auto 3D IMA of 11.0 (95% CI, 10.5-11.5). The HV group demonstrated significantly increased IMA compared with controls (P < .0001), with a mean DRR IMA of 15.4 (95% CI, 14.8-16.1), mean auto 2D of 17.8 (95% CI, 17.2-18.4), and mean auto 3D IMA of 16.8 (95% CI, 16.8-17.4). Measurements generated by the WBCT AI-based automatic measurement system for IMA demonstrated strong correlations with manual measurements, with near-perfect reproducibility. Further developments are warranted in order to make this tool more usable in daily practice, particularly with respect to its use in the presence of hardware in the foot. Level III, retrospective comparative study.

Quantifying lung and diaphragm morphology using radiographs in normative pediatric subjects, and predicting CT-derived lung volume.

To quantify the effect of age on two-dimensional (2D) radiographic lung and diaphragm morphology and determine if 2D radiographic lung measurements can be used to estimate computer tomography (CT)-derived lung volume in normative pediatric subjects. Digitally reconstructed radiographs (DRRs) were created using retrospective chest CT scans from 77 pediatric male and female subjects aged birth to 19 years. 2D lung and diaphragm measurements were made on the DRRs using custom MATLAB code, and Spearman correlations and exponential regression equations were used to relate 2D measurements with age. In addition, 3D lung volumes were segmented using CT scans, and power regression equations were fitted to predict each lung's CT-derived volume from 2D lung measurements. The coefficient of determination (R2 ) and standard error of the estimate (SEE) were used to assess the precision of the predictive equations with p < .05 indicating statistical significance. All 2D radiographic lung and diaphragm measurements showed statistically significant positive correlations with age (p < .01), including lung major axis (Spearman rho ≥ 0.90). Precise estimations of CT-derived lung volumes can be made using 2D lung measurements (R2 ≥ 0.95), including lung major axis (R2 ≥ 0.97). The reported pediatric age-specific reference data on 2D lung and diaphragm morphology and growth rates could be clinically used to identify lung and diaphragm pathologies during chest X-ray evaluations. The simple, precise, and clinically adaptable radiographic method for estimating CT-derived lung volumes may be used when pulmonary function tests are not readily available or difficult to perform.

A 2D/3D Non-rigid Registration Method for Lung Images Based on a Non-linear Correlation Between Displacement Vectors and Similarity Measures

This paper presents a novel 2D/3D non-rigid registration method for lung lesions tracking in image-guided diagnoses and treatments. Preoperative 3D lung CT volumes were obtained at a series of respiratory phases and important anatomical points were extracted. A CT volume was selected as reference volume and others were considered floating volumes. Displacement vectors of anatomical points were calculated using coherent point drift (CPD) and diffeomorphic-demon methods. For each CT volume, 2D digitally reconstructed radiographs (DRR) were generated by ray projection to simulate intraoperative 2D X-ray, and grayscale-based similarity measures of DRR between reference and floating volumes were calculated. A pulmonary respiration model was constructed using cubic polynomial, which represents the non-linear correlation between displacement vectors and similarity measures. During operation, the pulmonary respiration model used X-ray to obtain displacement vectors of important anatomical points; deformation fields were calculated from the displacement vectors through B-spline transformations. Finally, simulated lung CT volumes corresponding to intraoperative X-ray were generated by applying deformation fields to reference volume. The proposed method was compared to a state-of-the-art 2D/3D non-rigid registration method named respiratory-phase matching. Experimental results on two datasets respectively from a lung cancer patient and an ex vivo pig lung, showed mean registration accuracy of the proposed method was 1.57 ± 0.92 mm and 2.65 ± 1.35 mm; an improvement of 23% and 9% over the matching method. Moreover, mean lung structure overlaps were 0.91 and 0.88, comparable to matching method. The proposed method has potential in aiding lung intervention diagnoses and real-time lesion tracking.