Age Encoded Adversarial Learning for Pediatric CT Segmentation.

Organ segmentation from CT images is critical in the early diagnosis of diseases, progress monitoring, pre-operative planning, radiation therapy planning, and CT dose estimation. However, data limitation remains one of the main challenges in medical image segmentation tasks. This challenge is particularly huge in pediatric CT segmentation due to children's heightened sensitivity to radiation. In order to address this issue, we propose a novel segmentation framework with a built-in auxiliary classifier generative adversarial network (ACGAN) that conditions age, simultaneously generating additional features during training. The proposed conditional feature generation segmentation network (CFG-SegNet) was trained on a single loss function and used 2.5D segmentation batches. Our experiment was performed on a dataset with 359 subjects (180 male and 179 female) aged from 5 days to 16 years and a mean age of 7 years. CFG-SegNet achieved an average segmentation accuracy of 0.681 dice similarity coefficient (DSC) on the prostate, 0.619 DSC on the uterus, 0.912 DSC on the liver, and 0.832 DSC on the heart with four-fold cross-validation. We compared the segmentation accuracy of our proposed method with previously published U-Net results, and our network improved the segmentation accuracy by 2.7%, 2.6%, 2.8%, and 3.4% for the prostate, uterus, liver, and heart, respectively. The results indicate that our high-performing segmentation framework can more precisely segment organs when limited training images are available.

Dual energy CT reconstruction using the constrained one step spectral image reconstruction algorithm.

BACKGROUND: The constrained one-step spectral CT Image Reconstruction method (cOSSCIR) has been developed to estimate basis material maps directly from spectral CT data using a model of the polyenergetic x-ray transmissions and incorporating convex constraints into the inversion problem. This 'one-step' approach has been shown to stabilize the inversion in the case of photon-counting CT, and may provide similar benefits to dual-kV systems that utilize integrating detectors. Since the approach does not require the same rays be acquired for every spectral measurement, cOSSCIR can apply to dual energy protocols and systems used clinically, such as fast and slow kV switching systems and dual source scanning. PURPOSE: The purpose of this study is to investigate the use of cOSSCIR applied to dual-kV data, using both registered and unregistered spectral acquisitions, specifically slow and fast kV switching imaging protocols. For this application, cOSSCIR is investigated using inverse crime simulations and dual-kV experiments. This study is the first demonstration of cOSSCIR on the dual-kV reconstruction problem. METHODS: An integrating detector model was developed for the purpose of reconstructing dual-kV data, and an inverse crime study was used to validate the detector model within the cOSSCIR framework using a simulated pelvic phantom. Experiments were also used to evaluate cOSSCIR on the dual energy problem. Dual-kV data was obtained from a physical phantom containing analogs of adipose, bone, and liver tissues, with the aim of recovering the material coefficients in the bone and adipose basis material maps. cOSSCIR was applied to acquisitions where all rays performed both spectral measurements (registered) and fast and slow kV switching acquisitions (unregistered). cOSSCIR was also compared to two image-domain decomposition approaches, where image-domain methods are the conventional approach for decomposing unregistered spectral data. RESULTS: Simulations demonstrate the application of cOSSCIR to the dual-kV inversion problem by successfully recovering the material basis maps on ideal data, while further showing that unregistered data presents a more challenging inversion problem. In our experimental reconstructions, the recovered basis material coefficient errors were found to be less than 6.5% in the bone, adipose, and liver regions for both registered and unregistered protocols. Similarly, the errors were less than 4% in the 50 keV virtual mono-energetic images, and the recovered material decomposition vectors nearly overlap their corresponding ground-truth vectors. Additionally, a preliminary two material decomposition study of iodine quantification recovered an average concentration of 9.2 mg/mL from a 10 mg/mL experimental iodine analog. CONCLUSIONS: Using our integrating detector and spectral models, cOSCCIR is capable of accurately recovering material basis maps from dual-kV data for both registered and unregistered data. The material decomposition quantification compare favorably to the image domain approaches, and our results were not affected by the imaging protocol. Our results also suggest the extension of cOSSCIR to iodine quantification using two material decomposition.

Deep learning classification of deep ultraviolet fluorescence images toward intra-operative margin assessment in breast cancer.

Background: Breast-conserving surgery is aimed at removing all cancerous cells while minimizing the loss of healthy tissue. To ensure a balance between complete resection of cancer and preservation of healthy tissue, it is necessary to assess themargins of the removed specimen during the operation. Deep ultraviolet (DUV) fluorescence scanning microscopy provides rapid whole-surface imaging (WSI) of resected tissues with significant contrast between malignant and normal/benign tissue. Intra-operative margin assessment with DUV images would benefit from an automated breast cancer classification method. Methods: Deep learning has shown promising results in breast cancer classification, but the limited DUV image dataset presents the challenge of overfitting to train a robust network. To overcome this challenge, the DUV-WSI images are split into small patches, and features are extracted using a pre-trained convolutional neural network-afterward, a gradient-boosting tree trains on these features for patch-level classification. An ensemble learning approach merges patch-level classification results and regional importance to determine the margin status. An explainable artificial intelligence method calculates the regional importance values. Results: The proposed method's ability to determine the DUV WSI was high with 95% accuracy. The 100% sensitivity shows that the method can detect malignant cases efficiently. The method could also accurately localize areas that contain malignant or normal/benign tissue. Conclusion: The proposed method outperforms the standard deep learning classification methods on the DUV breast surgical samples. The results suggest that it can be used to improve classification performance and identify cancerous regions more effectively.

Constrained one-step material decomposition reconstruction of head CT data from a silicon photon-counting prototype.

BACKGROUND: Spectral CT material decomposition provides quantitative information but is challenged by the instability of the inversion into basis materials. We have previously proposed the constrained One-Step Spectral CT Image Reconstruction (cOSSCIR) algorithm to stabilize the material decomposition inversion by directly estimating basis material images from spectral CT data. cOSSCIR was previously investigated on phantom data. PURPOSE: This study investigates the performance of cOSSCIR using head CT datasets acquired on a clinical photon-counting CT (PCCT) prototype. This is the first investigation of cOSSCIR for large-scale, anatomically complex, clinical PCCT data. The cOSSCIR decomposition is preceded by a spectrum estimation and nonlinear counts correction calibration step to address nonideal detector effects. METHODS: Head CT data were acquired on an early prototype clinical PCCT system using an edge-on silicon detector with eight energy bins. Calibration data of a step wedge phantom were also acquired and used to train a spectral model to account for the source spectrum and detector spectral response, and also to train a nonlinear counts correction model to account for pulse pileup effects. The cOSSCIR algorithm optimized the bone and adipose basis images directly from the photon counts data, while placing a grouped total variation (TV) constraint on the basis images. For comparison, basis images were also reconstructed by a two-step projection-domain approach of Maximum Likelihood Estimation (MLE) for decomposing basis sinograms, followed by filtered backprojection (MLE + FBP) or a TV minimization algorithm (MLE + TVmin ) to reconstruct basis images. We hypothesize that the cOSSCIR approach will provide a more stable inversion into basis images compared to two-step approaches. To investigate this hypothesis, the noise standard deviation in bone and soft-tissue regions of interest (ROIs) in the reconstructed images were compared between cOSSCIR and the two-step methods for a range of regularization constraint settings. RESULTS: cOSSCIR reduced the noise standard deviation in the basis images by a factor of two to six compared to that of MLE + TVmin , when both algorithms were constrained to produce images with the same TV. The cOSSCIR images demonstrated qualitatively improved spatial resolution and depiction of fine anatomical detail. The MLE + TVmin algorithm resulted in lower noise standard deviation than cOSSCIR for the virtual monoenergetic images (VMIs) at higher energy levels and constraint settings, while the cOSSCIR VMIs resulted in lower noise standard deviation at lower energy levels and overall higher qualitative spatial resolution. There were no statistically significant differences in the mean values within the bone region of images reconstructed by the studied algorithms. There were statistically significant differences in the mean values within the soft-tissue region of the reconstructed images, with cOSSCIR producing mean values closer to the expected values. CONCLUSIONS: The cOSSCIR algorithm, combined with our previously proposed spectral model estimation and nonlinear counts correction method, successfully estimated bone and adipose basis images from high resolution, large-scale patient data from a clinical PCCT prototype. The cOSSCIR basis images were able to depict fine anatomical details with a factor of two to six reduction in noise standard deviation compared to that of the MLE + TVmin two-step approach.

Automated assessment of breast margins in deep ultraviolet fluorescence images using texture analysis.

Microscopy with ultraviolet surface excitation (MUSE) is increasingly studied for intraoperative assessment of tumor margins during breast-conserving surgery to reduce the re-excision rate. Here we report a two-step classification approach using texture analysis of MUSE images to automate the margin detection. A study dataset consisting of MUSE images from 66 human breast tissues was constructed for model training and validation. Features extracted using six texture analysis methods were investigated for tissue characterization, and a support vector machine was trained for binary classification of image patches within a full image based on selected feature subsets. A weighted majority voting strategy classified a sample as tumor or normal. Using the eight most predictive features ranked by the maximum relevance minimum redundancy and Laplacian scores methods has achieved a sample classification accuracy of 92.4% and 93.0%, respectively. Local binary pattern alone has achieved an accuracy of 90.3%.

Reduced Chest Computed Tomography Scan Length for Patients Positive for Coronavirus Disease 2019: Dose Reduction and Impact on Diagnostic Utility.

METHODS: This study used the Personalized Rapid Estimation of Dose in CT (PREDICT) tool to estimate patient-specific organ doses from CT image data. The PREDICT is a research tool that combines a linear Boltzmann transport equation solver for radiation dose map generation with deep learning algorithms for organ contouring. Computed tomography images from 74 subjects in the Medical Imaging Data Resource Center-RSNA International COVID-19 Open Radiology Database data set (chest CT of adult patients positive for COVID-19), which included expert annotations including "infectious opacities," were analyzed. First, the full z-scan length of the CT image data set was evaluated. Next, the z-scan length was reduced from the left hemidiaphragm to the top of the aortic arch. Generic dose reduction based on dose length product (DLP) and patient-specific organ dose reductions were calculated. The percentage of infectious opacities excluded from the reduced z-scan length was used to quantify the effect on diagnostic utility. RESULTS: Generic dose reduction, based on DLP, was 69%. The organ dose reduction ranged from approximately equal to 18% (breasts) to approximately equal to 64% (bone surface and bone marrow). On average, 12.4% of the infectious opacities were not included in the reduced z-coverage, per patient, of which 5.1% were above the top of the arch and 7.5% below the left hemidiaphragm. CONCLUSIONS: Limiting z-scan length of chest CTs reduced radiation dose without significantly compromising diagnostic utility in COVID-19 patients. The PREDICT demonstrated that patient-specific organ dose reductions varied from generic dose reduction based on DLP.

Addressing CT metal artifacts using photon-counting detectors and one-step spectral CT image reconstruction.

PURPOSE: The constrained one-step spectral CT image reconstruction (cOSSCIR) algorithm with a nonconvex alternating direction method of multipliers optimizer is proposed for addressing computed tomography (CT) metal artifacts caused by beam hardening, noise, and photon starvation. The quantitative performance of cOSSCIR is investigated through a series of photon-counting CT simulations. METHODS: cOSSCIR directly estimates basis material maps from photon-counting data using a physics-based forward model that accounts for beam hardening. The cOSSCIR optimization framework places constraints on the basis maps, which we hypothesize will stabilize the decomposition and reduce streaks caused by noise and photon starvation. Another advantage of cOSSCIR is that the spectral data need not be registered, so that a ray can be used even if some energy window measurements are unavailable. Photon-counting CT acquisitions of a virtual pelvic phantom with low-contrast soft tissue texture and bilateral hip prostheses were simulated. Bone and water basis maps were estimated using the cOSSCIR algorithm and combined to form a virtual monoenergetic image for the evaluation of metal artifacts. The cOSSCIR images were compared to a "two-step" decomposition approach that first estimated basis sinograms using a maximum likelihood algorithm and then reconstructed basis maps using an iterative total variation constrained least-squares optimization (MLE+TV min $_{\text{min}}$ ). Images were also compared to a nonspectral TV min $_{\text{min}}$ reconstruction of the total number of counts detected for each ray with and without normalized metal artifact reduction (NMAR) applied. The simulated metal density was increased to investigate the effects of increasing photon starvation. The quantitative error and standard deviation in regions of the phantom were compared across the investigated algorithms. The ability of cOSSCIR to reproduce the soft-tissue texture, while reducing metal artifacts, was quantitatively evaluated. RESULTS: Noiseless simulations demonstrated the convergence of the cOSSCIR and MLE+TV min $_{\text{min}}$ algorithms to the correct basis maps in the presence of beam-hardening effects. When noise was simulated, cOSSCIR demonstrated a quantitative error of -1 HU, compared to 2 HU error for the MLE+TV min $_{\text{min}}$ algorithm and -154 HU error for the nonspectral TV min $_{\text{min}}$ +NMAR algorithm. For the cOSSCIR algorithm, the standard deviation in the central iodine region of interest was 20 HU, compared to 299 HU for the MLE+TV min $_{\text{min}}$ algorithm, 41 HU for the MLE+TV min $_{\text{min}}$ +Mask algorithm that excluded rays through metal, and 55 HU for the nonspectral TV min $_{\text{min}}$ +NMAR algorithm. Increasing levels of photon starvation did not impact the bias or standard deviation of the cOSSCIR images. cOSSCIR was able to reproduce the soft-tissue texture when an appropriate regularization constraint value was selected. CONCLUSIONS: By directly inverting photon-counting CT data into basis maps using an accurate physics-based forward model and a constrained optimization algorithm, cOSSCIR avoids metal artifacts due to beam hardening, noise, and photon starvation. The cOSSCIR algorithm demonstrated improved stability and accuracy compared to a two-step method of decomposition followed by reconstruction.

Pediatric chest-abdomen-pelvis and abdomen-pelvis CT images with expert organ contours.

PURPOSE: Organ autosegmentation efforts to date have largely been focused on adult populations, due to limited availability of pediatric training data. Pediatric patients may present additional challenges for organ segmentation. This paper describes a dataset of 359 pediatric chest-abdomen-pelvis and abdomen-pelvis Computed Tomography (CT) images with expert contours of up to 29 anatomical organ structures to aid in the evaluation and development of autosegmentation algorithms for pediatric CT imaging. ACQUISITION AND VALIDATION METHODS: The dataset collection consists of axial CT images in Digital Imaging and Communications in Medicine (DICOM) format of 180 male and 179 female pediatric chest-abdomen-pelvis or abdomen-pelvis exams acquired from one of three CT scanners at Children's Wisconsin. The datasets represent random pediatric cases based upon routine clinical indications. Subjects ranged in age from 5 days to 16 years, with a mean age of 7 years. The CT acquisition, contrast, and reconstruction protocols varied across the scanner models and patients, with specifications available in the DICOM headers. Expert contours were manually labeled for up to 29 organ structures per subject. Not all contours are available for all subjects, due to limited field of view or unreliable contouring due to high noise. DATA FORMAT AND USAGE NOTES: The data are available on The Cancer Imaging Archive (TCIA_ (https://www.cancerimagingarchive.net/) under the collection Pediatric-CT-SEG. The axial CT image slices for each subject are available in DICOM format. The expert contours are stored in a single DICOM RTSTRUCT file for each subject. The contour names are listed in Table 2. POTENTIAL APPLICATIONS: This dataset will enable the evaluation and development of organ autosegmentation algorithms for pediatric populations, which exhibit variations in organ shape and size across age. Automated organ segmentation from CT images has numerous applications including radiation therapy, diagnostic tasks, surgical planning, and patient-specific organ dose estimation.

Validation of a deterministic linear Boltzmann transport equation solver for rapid CT dose computation using physical dose measurements in pediatric phantoms.

PURPOSE: The risk of inducing cancer to patients undergoing CT examinations has motivated efforts for CT dose estimation, monitoring, and reduction, especially among pediatric population. The method investigated in this study is Acuros CTD (Varian Medical Systems, Palo Alto, CA), a deterministic linear Boltzmann transport equation (LBTE) solver aimed at generating rapid and reliable dose maps of CT exams. By applying organ contours, organ doses can also be obtained, thus patient-specific organ dose estimates can be provided. This study experimentally validated Acuros against measurements performed on a clinical CT system using a range of physical pediatric anthropomorphic phantoms and acquisition protocols. METHODS: The study consisted of (1) the acquisition of dose measurements on a clinical CT scanner through thermoluminescent dosimeters (TLDs), and (2) the modeling in the Acuros platform of the measurement set up, which includes the modeling of the CT scanner and of the anthropomorphic phantoms. For the measurements, 1-year-old, 5-year-old, and 10-year-old anthropomorphic phantoms of the CIRS ATOM family were used. TLDs were placed in selected organ locations such as stomach, liver, lungs, and heart. The pediatric phantoms were scanned helically with the GE Discovery 750 HD clinical scanner for several examination protocols. For the simulations in Acuros, scanner-specific input, such as bowtie filters, overrange collimation, and tube current modulation schemes, were modeled. These scanner complexities were implemented by defining discretized X-ray beams whose spectral distribution, defined in Acuros by only six energy bins, varied across fan angle, cone angle, and slice position. The images generated during the CT acquisitions were used to create the geometrical models, by applying thresholding algorithms and assigning materials to the HU values. The TLDs were contoured in the phantom models as sensitive cylindrical volumes at the locations selected for dosimeters placement, to provide dose estimates, in terms of dose per unit photon. To compare measured doses with dose estimates, a calibration factor was derived from the CTDIvol displayed by the scanner, to account for the number of photons emitted by the X-ray tube during the procedure. RESULTS: The differences of the measured and estimated doses, in terms of absolute % errors, were within 13% for 153 TLD locations, with an error of 17% at the stomach for one study with the 10-year-old phantom. Root-mean-squared-errors (RMSE) across all TLD locations for all configurations were in the range of 3%-8%, with Acuros providing dose estimates in a time range of a few seconds up to 2 min. CONCLUSIONS: An overall good agreement between measurements and simulations was achieved, with average RMSE of 6% across all cases. The results demonstrate that Acuros can model a specific clinical scanner despite the required discretization in spatial and energy domains. The proposed deterministic tool has the potential to be part of a near real-time individualized dosimetry monitoring system for CT applications, providing patient-specific organ dose estimates.

Automated image processing pipeline for adaptive optics scanning light ophthalmoscopy.

To mitigate the substantial post-processing burden associated with adaptive optics scanning light ophthalmoscopy (AOSLO), we have developed an open-source, automated AOSLO image processing pipeline with both "live" and "full" modes. The live mode provides feedback during acquisition, while the full mode is intended to automatically integrate the copious disparate modules currently used in generating analyzable montages. The mean (±SD) lag between initiation and montage placement for the live pipeline was 54.6 ± 32.7s. The full pipeline reduced overall human operator time by 54.9 ± 28.4%, with no significant difference in resultant cone density metrics. The reduced overhead decreases both the technical burden and operating cost of AOSLO imaging, increasing overall clinical accessibility.

Experimental study of photon-counting CT neural network material decomposition under conditions of pulse pileup.

Purpose: We investigated the performance of a neural network (NN) material decomposition method under varying pileup conditions. Approach: Experiments were performed at tube current settings that provided count rates incident on the detector through air equal to 9%, 14%, 27%, 40%, and 54% of the maximum detector count rate. An NN was trained for each count-rate level using transmission measurements through known thicknesses of basis materials (PMMA and aluminum). The NN trained for each count-rate level was applied to x-ray transmission measurements through test materials and to CT data of a rod phantom. Material decomposition error was evaluated as the distance in basis material space between the estimated thicknesses and ground truth. Results: There was no clear trend between count-rate level and material decomposition error for all test materials except neoprene. As an example result, Teflon error was 0.33 cm at the 9% count-rate level and 0.12 cm at the 54% count-rate level for the x-ray transmission experiments. Decomposition error increased with count-rate level for the neoprene test case, with 0.65-cm error at 9% count-rate level and 1.14-cm error at the 54% count-rate level. In the CT study, material decomposition error decreased with increasing incident count rate. For example, the material decomposition error for Teflon was 0.089, 0.066, 0.054 at count-rate levels of 14%, 27%, and 40%, respectively. Conclusions: Results demonstrate over a range of incident count-rate levels that an NN trained at a specific count-rate level can learn the relationship between photon-counting spectral measurements and basis material thicknesses.

Rapid assessment of breast tumor margins using deep ultraviolet fluorescence scanning microscopy.

SIGNIFICANCE: Re-excision rates for women with invasive breast cancer undergoing breast conserving surgery (or lumpectomy) have decreased in the past decade but remain substantial. This is mainly due to the inability to assess the entire surface of an excised lumpectomy specimen efficiently and accurately during surgery. AIM: The goal of this study was to develop a deep-ultraviolet scanning fluorescence microscope (DUV-FSM) that can be used to accurately and rapidly detect cancer cells on the surface of excised breast tissue. APPROACH: A DUV-FSM was used to image the surfaces of 47 (31 malignant and 16 normal/benign) fresh breast tissue samples stained in propidium iodide and eosin Y solutions. A set of fluorescence images were obtained from each sample using low magnification (4 × ) and fully automated scanning. The images were stitched to form a color image. Three nonmedical evaluators were trained to interpret and assess the fluorescence images. Nuclear-cytoplasm ratio (N/C) was calculated and used for tissue classification. RESULTS: DUV-FSM images a breast sample with subcellular resolution at a speed of 1.0 min / cm2. Fluorescence images show excellent visual contrast in color, tissue texture, cell density, and shape between invasive carcinomas and their normal counterparts. Visual interpretation of fluorescence images by nonmedical evaluators was able to distinguish invasive carcinoma from normal samples with high sensitivity (97.62%) and specificity (92.86%). Using N/C alone was able to differentiate patch-level invasive carcinoma from normal breast tissues with reasonable sensitivity (81.5%) and specificity (78.5%). CONCLUSIONS: DUV-FSM achieved a good balance between imaging speed and spatial resolution with excellent contrast, which allows either visual or quantitative detection of invasive cancer cells on the surfaces of a breast surgical specimen.

Deterministic linear Boltzmann transport equation solver for patient-specific CT dose estimation: Comparison against a Monte Carlo benchmark for realistic scanner configurations and patient models.

PURPOSE: Epidemiological evidence suggests an increased risk of cancer related to computed tomography (CT) scans, with children exposed to greater risk. The purpose of this work is to test the reliability of a linear Boltzmann transport equation (LBTE) solver for rapid and patient-specific CT dose estimation. This includes building a flexible LBTE framework for modeling modern clinical CT scanners and to validate the resulting dose maps across a range of realistic scanner configurations and patient models. METHODS: In this study, computational tools were developed for modeling CT scanners, including a bowtie filter, overrange collimation, and tube current modulation. The LBTE solver requires discretization in the spatial, angular, and spectral dimensions, which may affect the accuracy of scanner modeling. To investigate these effects, this study evaluated the LBTE dose accuracy for different discretization parameters, scanner configurations, and patient models (male, female, adults, pediatric). The method used to validate the LBTE dose maps was the Monte Carlo code Geant4, which provided ground truth dose maps. LBTE simulations were implemented on a GeForce GTX 1080 graphic unit, while Geant4 was implemented on a distributed cluster of CPUs. RESULTS: The agreement between Geant4 and the LBTE solver quantifies the accuracy of the LBTE, which was similar across the different protocols and phantoms. The results suggest that 18 views per rotation provides sufficient accuracy, as no significant improvement in the accuracy was observed by increasing the number of projection views. Considering this discretization, the LBTE solver average simulation time was approximately 30 s. However, in the LBTE solver the phantom model was implemented with a lower voxel resolution with respect to Geant4, as it is limited by the memory of the GPU. Despite this discretization, the results showed a good agreement between the LBTE and Geant4, with root mean square error of the dose in organs of approximately 3.5% for most of the studied configurations. CONCLUSIONS: The LBTE solver is proposed as an alternative to Monte Carlo for patient-specific organ dose estimation. This study demonstrated accurate organ dose estimates for the rapid LBTE solver when considering realistic aspects of CT scanners and a range of phantom models. Future plans will combine the LBTE framework with deep learning autosegmentation algorithms to provide near real-time patient-specific organ dose estimation.

Experimental investigation of neural network estimator and transfer learning techniques for K-edge spectral CT imaging.

PURPOSE: Spectral computed tomography (CT) material decomposition algorithms require accurate physics-based models or empirically derived models. This study investigates a machine learning algorithm and transfer learning techniques for Spectral CT imaging of K-edge contrast agents using simulated and experimental measurements. METHODS: A feed forward multilayer perceptron was implemented and trained on data acquired using a step wedge phantom containing acrylic, aluminum, and gadolinium materials. The neural network estimator was evaluated by scanning a rod phantom with varying dilutions of gadolinium oxide nanoparticles and by scanning a rat leg specimen with injected nanoparticles on a bench-top photon-counting computed tomography system. The algorithm decomposed each spectral projection measurement into path lengths of acrylic and aluminum and mass lengths of gadolinium. Each basis material sinogram was reconstructed into basis material images using filtered backprojection. Machine learning techniques of data standardization, transfer learning from aggregated pixel data, and transfer learning from simulations were investigated to improve image quality. The algorithm was compared to a previously published empirical method based on a linear approximation and calibration error look-up tables. RESULTS: The combined transfer learning techniques did not improve quantification in the rod phantom and provided only a small qualitative improvement in ring artifacts. Transfer learning from aggregated pixel data and from simulations improved the qualitative image quality of the rat specimen, for which the calibration data were limited. Transfer learning from aggregated pixel data and simulations estimated 3.26, 6.26, and 12.45 mg/mL Gd concentrations compared to true 2.72, 5.44, and 10.88 mg/mL concentrations in the rod phantom. Additionally, the neural networks were able to separate the soft tissue, bone, and gadolinium nanoparticles of the ex vivo rat leg specimen into the different basis images. CONCLUSIONS: The results demonstrate that empirical K-edge imaging from calibration measurements using machine learning and transfer learning is possible without explicit models of material attenuations, incident spectra, or the detector response.

Estimation of changing gross tumor volume from longitudinal CTs during radiation therapy delivery based on a texture analysis with classifier algorithms: a proof-of-concept study.

BACKGROUND: Adaptive radiation therapy (ART) is moving into the clinic rapidly. Capability of delineating the tumor change as a result of treatment response during treatment delivery is essential for ART. During image-guided radiation therapy (IGRT), a CT or cone-beam CT is taken at the time of daily setup and the tumor is not visible by eye in regions of soft tissue due to low contrast. The scope of this paper is to develop a method using a classifier trained on non-contrast CT textures, to estimate the gross tumor volume (GTV) of the day (GTVd) from daily (longitudinal) CTs acquired during the course of IGRT when the tumor is not visible. METHODS: CT textures from daily diagnostic-quality CTs routinely acquired during IGRT using an in-room CT were analyzed. Pretreatment GTV was delineated from pre-RT diagnostic images and populated to the first daily CT. Maps of first-order textures (mean, SD, entropy, skewness and kurtosis) and short-range second-order textures were created from the first daily CT. The classifier was trained to sort voxels into GTV and surrounding tissue on subsequent daily CTs over the course of RT. Optimum combinations of textures was defined by repeating the training process with all possible texture combinations. The trained classifier was used to identify voxels belonging to the GTVd, based on the CT of the day. Posttreatment GTV delineated from the post-RT follow-up images was populated to the last daily CT and used to validate the last GTVd delineated by the classifier. To demonstrate the concept, the method was described using three representative treatment sites, e.g., lung, breast and pancreatic tumors. RESULTS: Comparing the classifier map generated from a new CT to the initial training CT, the dice coefficient (DC) for GTV in lung is 83% on the eighth treatment and 84% on the last. The DC for the breast GTV is 56% mid-treatment and 65% at the last treatment. In the case of the pancreas with the least in organ tissue contrast, the DC for 4 cases ranges from 21% to 77% for the last treatment compared with the post-RT diagnostic CT. The Housdorff distance (HD) ranged from 2.9 to 5.9 mm with the mean GTV RECIST dimension of 22.75 mm long by 14.7 mm short. CONCLUSIONS: It is feasible to estimate the general region of the GTV of the day from the daily CT acquired during RT, based on CT textures, using a trained voxel classifier algorithm. The obtained GTV may be used as a starting point for an accurate GTV delineation in online adaptive replanning. Further study with larger patient datasets is required to improve the robustness of the algorithms.

The Utility of Frame Averaging for Automated Algorithms in Analyzing Retinal Vascular Biomarkers in AngioVue OCTA.

PURPOSE: This study proposes an optical coherence tomography angiography (OCTA) frame-averaging method and investigates the effects of the number of frames acquired and averaged on metrics quantifying the foveal avascular zone (FAZ), vessel morphology, and parafoveal intercapillary area (PICA). METHODS: Ten OCTA frames were acquired for each of the 19 subjects without known retinal disease using the AngioVue OCTA system. For each subject, acquired frames were ranked by an image quality metric. A subset of frames was then registered and averaged. The effects of the number of frames acquired and averaged on FAZ segmentation and metrics of FAZ geometry, vessel morphology, and PICA were analyzed. RESULTS: Frame averaging increased the accuracy of the automatically segmented FAZ region; for example, the absolute error in FAZ area decreased from 0.026 mm2 (1 frame) to 0.005 mm2 (5 frames). Averaging multiple frames exponentially decreased the estimated number of vessel endpoints and increased the average vessel length with a 32% decrease in number of endpoints and 14% increase in average vessel length when averaging five frames compared with one. Frame averaging also improved the precision of PICA estimates. CONCLUSIONS: Averaging multiple OCTA frames using the Optovue AngioVue system reduced error in FAZ segmentation and improved the robustness of OCTA vessel morphology and perfusion metrics. The study demonstrated limited benefit in acquiring and averaging more than five frames. TRANSLATIONAL RELEVANCE: Averaging multiple OCTA frames improved the robustness of OCTA foveal biomarkers with limited benefit when averaging more than five frames.

CT automated exposure control using a generalized detectability index.

PURPOSE: Identifying an appropriate tube current setting can be challenging when using iterative reconstruction due to the varying relationship between spatial resolution, contrast, noise, and dose across different algorithms. This study developed and investigated the application of a generalized detectability index ( d gen ' ) to determine the noise parameter to input to existing automated exposure control (AEC) systems to provide consistent image quality (IQ) across different reconstruction approaches. METHODS: This study proposes a task-based automated exposure control (AEC) method using a generalized detectability index ( d gen ' ). The proposed method leverages existing AEC methods that are based on a prescribed noise level. The generalized d gen ' metric is calculated using lookup tables of task-based modulation transfer function (MTF) and noise power spectrum (NPS). To generate the lookup tables, the American College of Radiology CT accreditation phantom was scanned on a multidetector CT scanner (Revolution CT, GE Healthcare) at 120 kV and tube current varied manually from 20 to 240 mAs. Images were reconstructed using a reference reconstruction algorithm and four levels of an in-house iterative reconstruction algorithm with different regularization strengths (IR1-IR4). The task-based MTF and NPS were estimated from the measured images to create lookup tables of scaling factors that convert between d gen ' and noise standard deviation. The performance of the proposed d gen ' -AEC method in providing a desired IQ level over a range of iterative reconstruction algorithms was evaluated using the American College of Radiology (ACR) phantom with elliptical shell and using a human reader evaluation on anthropomorphic phantom images. RESULTS: The study of the ACR phantom with elliptical shell demonstrated reasonable agreement between the d gen ' predicted by the lookup table and d ' measured in the images, with a mean absolute error of 15% across all dose levels and maximum error of 45% at the lowest dose level with the elliptical shell. For the anthropomorphic phantom study, the mean reader scores for images resulting from the d gen ' -AEC method were 3.3 (reference image), 3.5 (IR1), 3.6 (IR2), 3.5 (IR3), and 2.2 (IR4). When using the d gen ' -AEC method, the observers' IQ scores for the reference reconstruction were statistical equivalent to the scores for IR1, IR2, and IR3 iterative reconstructions (P > 0.35). The d gen ' -AEC method achieved this equivalent IQ at lower dose for the IR scans compared to the reference scans. CONCLUSIONS: A novel AEC method, based on a generalized detectability index, was investigated. The proposed method can be used with some existing AEC systems to derive the tube current profile for iterative reconstruction algorithms. The results provide preliminary evidence that the proposed d gen ' -AEC can produce similar IQ across different iterative reconstruction approaches at different dose levels.

A fast, linear Boltzmann transport equation solver for computed tomography dose calculation (Acuros CTD).

PURPOSE: To improve dose reporting of CT scans, patient-specific organ doses are highly desired. However, estimating the dose distribution in a fast and accurate manner remains challenging, despite advances in Monte Carlo methods. In this work, we present an alternative method that deterministically solves the linear Boltzmann transport equation (LBTE), which governs the behavior of x-ray photon transport through an object. METHODS: Our deterministic solver for CT dose (Acuros CTD) is based on the same approach used to estimate scatter in projection images of a CT scan (Acuros CTS). A deterministic method is used to compute photon fluence within the object, which is then converted to deposited energy by multiplying by known, material-specific conversion factors. To benchmark Acuros CTD, we used the AAPM Task Group 195 test for CT dose, which models an axial, fan beam scan (10 mm thick beam) and calculates energy deposited in each organ of an anthropomorphic phantom. We also validated our own Monte Carlo implementation of Geant4 to use as a reference to compare Acuros against for other common geometries like an axial, cone beam scan (160 mm thick beam) and a helical scan (40 mm thick beam with table motion for a pitch of 1). RESULTS: For the fan beam scan, Acuros CTD accurately estimated organ dose, with a maximum error of 2.7% and RMSE of 1.4% when excluding organs with <0.1% of the total energy deposited. The cone beam and helical scans yielded similar levels of accuracy compared to Geant4. Increasing the number of source positions beyond 18 or decreasing the voxel size below 5 × 5 × 5 mm3 provided marginal improvement to the accuracy for the cone beam scan but came at the expense of increased run time. Across the different scan geometries, run time of Acuros CTD ranged from 8 to 23 s. CONCLUSIONS: In this digital phantom study, a deterministic LBTE solver was capable of fast and accurate organ dose estimates.

Estimating the spectrum in computed tomography via Kullback-Leibler divergence constrained optimization.

PURPOSE: We study the problem of spectrum estimation from transmission data of a known phantom. The goal is to reconstruct an x-ray spectrum that can accurately model the x-ray transmission curves and reflects a realistic shape of the typical energy spectra of the CT system. METHODS: Spectrum estimation is posed as an optimization problem with x-ray spectrum as unknown variables, and a Kullback-Leibler (KL)-divergence constraint is employed to incorporate prior knowledge of the spectrum and enhance numerical stability of the estimation process. The formulated constrained optimization problem is convex and can be solved efficiently by use of the exponentiated-gradient (EG) algorithm. We demonstrate the effectiveness of the proposed approach on the simulated and experimental data. The comparison to the expectation-maximization (EM) method is also discussed. RESULTS: In simulations, the proposed algorithm is seen to yield x-ray spectra that closely match the ground truth and represent the attenuation process of x-ray photons in materials, both included and not included in the estimation process. In experiments, the calculated transmission curve is in good agreement with the measured transmission curve, and the estimated spectra exhibits physically realistic looking shapes. The results further show the comparable performance between the proposed optimization-based approach and EM. CONCLUSIONS: Our formulation of a constrained optimization provides an interpretable and flexible framework for spectrum estimation. Moreover, a KL-divergence constraint can include a prior spectrum and appears to capture important features of x-ray spectrum, allowing accurate and robust estimation of x-ray spectrum in CT imaging.

Application of fractal dimension for quantifying noise texture in computed tomography images.

PURPOSE: Evaluation of noise texture information in CT images is important for assessing image quality. Noise texture is often quantified by the noise power spectrum (NPS), which requires numerous image realizations to estimate. This study evaluated fractal dimension for quantifying noise texture as a scalar metric that can potentially be estimated using one image realization. METHODS: The American College of Radiology CT accreditation phantom (ACR) was scanned on a clinical scanner (Discovery CT750, GE Healthcare) at 120 kV and 25 and 90 mAs. Images were reconstructed using filtered back projection (FBP/ASIR 0%) with varying reconstruction kernels: Soft, Standard, Detail, Chest, Lung, Bone, and Edge. For each kernel, images were also reconstructed using ASIR 50% and ASIR 100% iterative reconstruction (IR) methods. Fractal dimension was estimated using the differential box-counting algorithm applied to images of the uniform section of ACR phantom. The two-dimensional Noise Power Spectrum (NPS) and one-dimensional-radially averaged NPS were estimated using established techniques. By changing the radiation dose, the effect of noise magnitude on fractal dimension was evaluated. The Spearman correlation between the fractal dimension and the frequency of the NPS peak was calculated. The number of images required to reliably estimate fractal dimension was determined and compared to the number of images required to estimate the NPS-peak frequency. The effect of Region of Interest (ROI) size on fractal dimension estimation was evaluated. Feasibility of estimating fractal dimension in an anthropomorphic phantom and clinical image was also investigated, with the resulting fractal dimension compared to that estimated within the uniform section of the ACR phantom. RESULTS: Fractal dimension was strongly correlated with the frequency of the peak of the radially averaged NPS curve, having a Spearman rank-order coefficient of 0.98 (P-value < 0.01) for ASIR 0%. The mean fractal dimension at ASIR 0% was 2.49 (Soft), 2.51 (Standard), 2.52 (Detail), 2.57 (Chest), 2.61 (Lung), 2.66 (Bone), and 2.7 (Edge). A reduction in fractal dimension was observed with increasing ASIR levels for all investigated reconstruction kernels. Fractal dimension was found to be independent of noise magnitude. Fractal dimension was successfully estimated from four ROIs of size 64 × 64 pixels or one ROI of 128 × 128 pixels. Fractal dimension was found to be sensitive to non-noise structures in the image, such as ring artifacts and anatomical structure. Fractal dimension estimated within a uniform region of an anthropomorphic phantom and clinical head image matched that estimated within the ACR phantom for filtered back projection reconstruction. CONCLUSIONS: Fractal dimension correlated with the NPS-peak frequency and was independent of noise magnitude, suggesting that the scalar metric of fractal dimension can be used to quantify the change in noise texture across reconstruction approaches. Results demonstrated that fractal dimension can be estimated from four, 64 × 64-pixel ROIs or one 128 × 128 ROI within a head CT image, which may make it amenable for quantifying noise texture within clinical images.