Data Extrapolation From Learned Prior Images for Truncation Correction in Computed Tomography.

Data truncation is a common problem in computed tomography (CT). Truncation causes cupping artifacts inside the field-of-view (FOV) and anatomical structures missing outside the FOV. Deep learning has achieved impressive results in CT reconstruction from limited data. However, its robustness is still a concern for clinical applications. Although the image quality of learning-based compensation schemes may be inadequate for clinical diagnosis, they can provide prior information for more accurate extrapolation than conventional heuristic extrapolation methods. With extrapolated projection, a conventional image reconstruction algorithm can be applied to obtain a final reconstruction. In this work, a general plug-and-play (PnP) method for truncation correction is proposed based on this idea, where various deep learning methods and conventional reconstruction algorithms can be plugged in. Such a PnP method integrates data consistency for measured data and learned prior image information for truncated data. This shows to have better robustness and interpretability than deep learning only. To demonstrate the efficacy of the proposed PnP method, two state-of-the-art deep learning methods, FBPConvNet and Pix2pixGAN, are investigated for truncation correction in cone-beam CT in noise-free and noisy cases. Their robustness is evaluated by showing false negative and false positive lesion cases. With our proposed PnP method, false lesion structures are corrected for both deep learning methods. For FBPConvNet, the root-mean-square error (RMSE) inside the FOV can be improved from 92HU to around 30HU by PnP in the noisy case. Pix2pixGAN solely achieves better image quality than FBPConvNet solely for truncation correction in general. PnP further improves the RMSE inside the FOV from 42HU to around 27HU for Pix2pixGAN. The efficacy of PnP is also demonstrated on real clinical head data.

Traditional machine learning for limited angle tomography.

PURPOSE: The application of traditional machine learning techniques, in the form of regression models based on conventional, "hand-crafted" features, to artifact reduction in limited angle tomography is investigated. METHODS: Mean-variation-median (MVM), Laplacian, Hessian, and shift-variant data loss (SVDL) features are extracted from the images reconstructed from limited angle data. The regression models linear regression (LR), multilayer perceptron (MLP), and reduced-error pruning tree (REPTree) are applied to predict artifact images. RESULTS: REPTree learns artifacts best and reaches the smallest root-mean-square error (RMSE) of 29 HU for the Shepp-Logan phantom in a parallel-beam study. Further experiments demonstrate that the MVM and Hessian features complement each other, whereas the Laplacian feature is redundant in the presence of MVM. In fan-beam, the SVDL features are also beneficial. A preliminary experiment on clinical data in a fan-beam study demonstrates that REPTree can reduce some artifacts for clinical data. However, it is not sufficient as a lot of incorrect pixel intensities still remain in the estimated reconstruction images. CONCLUSION: REPTree has the best performance on learning artifacts in limited angle tomography compared with LR and MLP. The features of MVM, Hessian, and SVDL are beneficial for artifact prediction in limited angle tomography. Preliminary experiments on clinical data suggest that the investigation on more features is necessary for clinical applications of REPTree.

Determinants of image quality of rotational angiography for on-line assessment of frame geometry after transcatheter aortic valve implantation.

To study the determinants of image quality of rotational angiography using dedicated research prototype software for motion compensation without rapid ventricular pacing after the implantation of four commercially available catheter-based valves. Prospective observational study including 179 consecutive patients who underwent transcatheter aortic valve implantation (TAVI) with either the Medtronic CoreValve (MCS), Edward-SAPIEN Valve (ESV), Boston Sadra Lotus (BSL) or Saint-Jude Portico Valve (SJP) in whom rotational angiography (R-angio) with motion compensation 3D image reconstruction was performed. Image quality was evaluated from grade 1 (excellent image quality) to grade 5 (strongly degraded). Distinction was made between good (grades 1, 2) and poor image quality (grades 3-5). Clinical (gender, body mass index, Agatston score, heart rate and rhythm, artifacts), procedural (valve type) and technical variables (isocentricity) were related with the image quality assessment. Image quality was good in 128 (72 %) and poor in 51 (28 %) patients. By univariable analysis only valve type (BSL) and the presence of an artefact negatively affected image quality. By multivariate analysis (in which BMI was forced into the model) BSL valve (Odds 3.5, 95 % CI [1.3-9.6], p = 0.02), presence of an artifact (Odds 2.5, 95 % CI [1.2-5.4], p = 0.02) and BMI (Odds 1.1, 95 % CI [1.0-1.2], p = 0.04) were independent predictors of poor image quality. Rotational angiography with motion compensation 3D image reconstruction using a dedicated research prototype software offers good image quality for the evaluation of frame geometry after TAVI in the majority of patients. Valve type, presence of artifacts and higher BMI negatively affect image quality.

Diferencias en geometría entre válvulas percutáneas expandibles con balón y autoexpandibles y su relación con la insuficiencia aórtica / Differences in frame geometry between balloon-expandable and self-expanding transcatheter heart valves and association with aortic regurgitation

Introducción y objetivos: Se sabe que los factores relacionados con el paciente y con la intervención se asocian con insuficiencia aórtica después de un implante percutáneo de válvula aórtica. No obstante, también puede causarla una interacción específica entre el dispositivo y el huésped como consecuencia de las propiedades biomecánicas de las válvulas, con independencia de los factores clínicos. El objetivo de este estudio es esclarecer el papel de la geometría de la válvula en la aparición de insuficiencia aórtica después del implante de las válvulas Medtronic CoreValve® y Edwards SAPIEN®. Métodos: Se llevó a cabo un estudio observacional que incluyó a 134 pacientes tratados con implante percutáneo de válvula aórtica empleando las válvulas Medtronic CoreValve® y Edwards SAPIEN®. El análisis geométrico se realizó en tres niveles predefinidos de ambas válvulas mediante angiografía rotacional con compensación de movimiento usando un programa informático específicamente desarrollado para este fin. Se estableció una distinción entre los pacientes con insuficiencia aórtica nula o leve y los pacientes con insuficiencia aórtica moderada o grave según la ecocardiografía. Resultados: Las características basales eran similares con ambas válvulas. A pesar del mayor uso de predilatación en el grupo de CoreValve® (el 95,2 frente al 82,0%; p = 0,012), el mayor exceso de tamaño de prótesis/anillo aórtico (perímetro, el 114 ± 7% frente al 103 ± 7%; p < 0,001) y la misma profundidad de implante (seno no coronario, 7 ± 4 frente a 8 ± 2 mm; seno coronario izquierdo, 8 ± 4 frente a 8 ± 2 mm), esta válvula tuvo menos expansión (el 83 ± 7% frente al 92 ± 4%; p < 0,001) y fue más excéntrica (el 82 ± 8% frente al 95 ± 3%; p < 0,001) que la válvula Edwards SAPIEN®, también tras introducir un ajuste de la excentricidad respecto a la excentricidad del anillo valvular del paciente (el 4 ± 13% frente al 21 ± 11%; p < 0,001). La excentricidad y la excentricidad ajustada se asociaron con insuficiencia aórtica moderada o grave. Conclusiones: Independientemente de los factores relacionados con el paciente y con la intervención, existe una interacción entre dispositivo y huésped que es específica del dispositivo y explica la insuficiencia aórtica existente después de un implante percutáneo de válvula aórtica (AU)

Introduction and objectives: Patient- and procedure-related factors are known to be associated with aortic regurgitation after transcatheter aortic valve implantation. Nevertheless, this entity may also be caused by a specific device-host interaction due to the biomechanical properties of the valves, independently of clinical factors. We sought to elucidate the role of frame geometry in the occurrence of aortic regurgitation after Medtronic CoreValve and Edwards SAPIEN valve implantation. Methods: We conducted an observational study encompassing 134 patients undergoing transcatheter aortic valve implantation with the Medtronic CoreValve and Edwards SAPIEN valve. Frame analysis was performed at 3 predefined levels of both valves by rotational angiography using dedicated motion compensation software. A distinction was made between patients with no-to-mild and moderate-to-severe aortic regurgitation by echocardiography. Results: Baseline characteristics were similar between the 2 valves. Despite greater use of predilation in the CoreValve (95.2% vs 82.0%; P = .012), more oversizing (perimeter, 114 ± 7% vs 103 ± 7%;P < .001), and the same depth of implantation (noncoronary sinus, 7 ± 4 vs 8 ± 2 mm; left coronary sinus, 8 ± 4 vs 8 ± 2 mm), it was less expanded and more eccentric than the Edwards SAPIEN (83 ± 7% vs 92 ± 4%; P < .001 and 82 ± 8% vs 95 ± 3%; P < .001, respectively) and when eccentricity was adjusted for the patient's annulus eccentricity (4 ± 13% vs 21 ± 11%;P < .001). Eccentricity and adjusted eccentricity were associated with moderate-to-severe aortic regurgitation. Conclusions: Independently of patient- and procedure-related factors, there is a device-specific device-host interaction that explains aortic regurgitation after transcatheter aortic valve implantation (AU)

Differences in Frame Geometry Between Balloon-expandable and Self-expanding Transcatheter Heart Valves and Association With Aortic Regurgitation.

INTRODUCTION AND OBJECTIVES: Patient- and procedure-related factors are known to be associated with aortic regurgitation after transcatheter aortic valve implantation. Nevertheless, this entity may also be caused by a specific device-host interaction due to the biomechanical properties of the valves, independently of clinical factors. We sought to elucidate the role of frame geometry in the occurrence of aortic regurgitation after Medtronic CoreValve and Edwards SAPIEN valve implantation. METHODS: We conducted an observational study encompassing 134 patients undergoing transcatheter aortic valve implantation with the Medtronic CoreValve and Edwards SAPIEN valve. Frame analysis was performed at 3 predefined levels of both valves by rotational angiography using dedicated motion compensation software. A distinction was made between patients with no-to-mild and moderate-to-severe aortic regurgitation by echocardiography. RESULTS: Baseline characteristics were similar between the 2 valves. Despite greater use of predilation in the CoreValve (95.2% vs 82.0%; P=.012), more oversizing (perimeter, 114±7% vs 103±7%; P<.001), and the same depth of implantation (noncoronary sinus, 7±4 vs 8±2mm; left coronary sinus, 8±4 vs 8±2mm), it was less expanded and more eccentric than the Edwards SAPIEN (83±7% vs 92±4%; P<.001 and 82±8% vs 95±3%; P<.001, respectively) and when eccentricity was adjusted for the patient's annulus eccentricity (4±13% vs 21±11%; P<.001). Eccentricity and adjusted eccentricity were associated with moderate-to-severe aortic regurgitation. CONCLUSIONS: Independently of patient- and procedure-related factors, there is a device-specific device-host interaction that explains aortic regurgitation after transcatheter aortic valve implantation.

Does frame geometry play a role in aortic regurgitation after Medtronic CoreValve implantation?

AIMS: Aortic regurgitation (AR) after Medtronic CoreValve System (MCS) implantation may be explained by patient-, operator- and procedure-related factors. We sought to explore if frame geometry, as a result of a specific device-host interaction, contributes to AR. METHODS AND RESULTS: Using rotational angiography with dedicated motion compensation, we assessed valve frame geometry in 84 patients who underwent TAVI with the MCS. Aortic regurgitation was assessed by angiography (n=84, Sellers) and echocardiography at discharge (n=72, VARC-2). Twenty-two patients (26%) had AR grade ≥2 using contrast angiography, and 17 (24%) by echocardiography. Balloon predilatation and sizing and depth of implantation did not differ between the two groups. Despite more frequent balloon post-dilatation in patients with AR (40.9 vs. 9.7%, p=0.001), the frame was more elliptical at its nadir relative to the patient's annulus (6±13 vs. -1±11%, p=0.046) and occurred in a larger proportion of patients (61.9 vs. 26.8%, p=0.004). Although the Agatston score and the eccentricity of the MCS frame relative to the annulus were independent determinants of AR (odds ratio: 1.635 [1.151-2.324], p=0.006, and 4.204 [1.237-14.290], p=0.021), there was a weak association between the Agatston score and the adjusted eccentricity (Spearman's rank correlation coefficient =-0.24, p=0.046). CONCLUSIONS: These findings indicate that AR can be explained by a specific device-host interaction which can only partially be explained by the calcium load of the aortic root.

Rotational angiography with motion compensation: first-in-man use for the 3D evaluation of transcatheter valve prostheses.

AIMS: We evaluated a novel motion-compensating 3D reconstruction technique applied to rotational angiography (R-angio) which produces MSCT-like images for evaluation of implanted TAVI prostheses without requiring rapid pacing. METHODS AND RESULTS: Fifty-one consecutive patients were retrospectively identified who were evaluated with rotational angiography (R-angio) using the Siemens Artis zee angiographic C-arm system after TAVI with a Medtronic CoreValve prosthesis. A novel 3D image reconstruction technique was applied which corrects for cardiac motion. CoreValve frame geometry was evaluated according to the same protocol for MSCT and R-angio at the level of: 1) the inflow, 2) the nadirs, 3) central coaptation, and 4) the commissures. The native aortic annulus dimensions were measured at the nadirs of the three leaflets. Sizing ratio, prosthesis expansion and frame ellipticity were assessed. Good quality 3D reconstructions were obtained in 43 patients (84%) and failure was predictable prior to reconstruction in six of the other seven patients (superposition of radiographically dense object n=4, obesity n=2). Prosthesis inflow ellipticity and expansion were correlated with implantation depth (respectively r=-0.46, p<0.01, and r=0.61, p<0.001). Aortic regurgitation grade ≥2 was associated with greater prosthesis ellipticity at the level of central coaptation (median [25th-75th percentile]: 1.15 [1.10-1.20] vs. 1.08 [1.06-1.12], p=0.009). The inter-observer, inter-modality (MSCT, R-angio) variability in measurement at the level of coaptation for minimum diameter, maximum diameter and area were all low (respectively, mean ±SD:1.2% ±1.2; 1.7% ±1.8 and 2.0% ±1.3). CONCLUSIONS: R-angio with motion-compensated reconstruction offers new possibilities for evaluation of the post-implantation geometry of percutaneous structural heart prostheses and the potential clinical effects.

Effect of body mass index on the image quality of rotational angiography without rapid pacing for planning of transcatheter aortic valve implantation: a comparison with multislice computed tomography.

AIMS: To evaluate the feasibility of procedural planning for transcatheter aortic valve implantation (TAVI) using rotational angiography (R-angio) by comparison with multislice computed tomography (MSCT) and to investigate determinants of the image quality of R-angio. METHODS AND RESULTS: Patients who underwent R-angio of the left ventricle and cardiac MSCT were eligible. R-angio acquisition was performed during contrast injection through a 6F pigtail catheter positioned in the left ventricle. On 3D R-angio and MSCT data sets, diameter measurements were made on short-axis images at the level of the aortic annulus (D(perimeter), D(area)), ascending aorta, sino-tubular junction (ST-junction), and the sinus of Valsalva. At the level of the aortic annulus, diagnostic image quality was obtained in 49 of 56 patients. In all patients with a body mass index (BMI) < 29 kg/m(2), image quality was acceptable whether or not rapid pacing was used. In patients with BMI ≥ 29 kg/m(2), the image quality was poor in 1 of 9 (11%) who were rapidly paced compared with 6 of 12 (50%) who were not. The correlation between R-angio and MSCT measurements was high for aortic annulus D(perimeter), D(area), ST-junction, Valsalva sinus, and ascending aorta (respectively, R = 0.90, 0.90, 0.91, 0.92, and 0.89). The correlations improved further when the analysis was limited to patients with a BMI < 29 kg/m(2) (respectively, 0.92, 0.92, 0.92, 0.92, and 0.93). CONCLUSION: R-angio of the left ventricle allows precise measurement of the aortic root and annulus and was feasible for sizing at the time of TAVI. Diagnostic image quality was obtained without rapid pacing in all patients with a BMI < 29 kg/m(2).

Automatic 3D motion estimation of left ventricle from C-arm rotational angiocardiography using a prior motion model and learning based boundary detector.

Compared to pre-operative imaging modalities, it is more convenient to estimate the current cardiac physiological status from C-arm angiocardiography since C-arm is a widely used intra-operative imaging modality to guide many cardiac interventions. The 3D shape and motion of the left ventricle (LV) estimated from rotational angiocardiography provide important cardiac function measurements, e.g., ejection fraction and myocardium motion dyssynchrony. However, automatic estimation of the 3D LV motion is difficult since all anatomical structures overlap on the 2D X-ray projections and the nearby confounding strong image boundaries (e.g., pericardium) often cause ambiguities to LV endocardium boundary detection. In this paper, a new framework is proposed to overcome the aforementioned difficulties: (1) A new learning-based boundary detector is developed by training a boosting boundary classifier combined with the principal component analysis of a local image patch; (2) The prior LV motion model is learned from a set of dynamic cardiac computed tomography (CT) sequences to provide a good initial estimate of the 3D LV shape of different cardiac phases; (3) The 3D motion trajectory is learned for each mesh point; (4) All these components are integrated into a multi-surface graph optimization method to extract the globally coherent motion. The method is tested on seven patient scans, showing significant improvement on the ambiguous boundary cases with a detection accuracy of 2.87 +/- 1.00 mm on LV endocardium boundary delineation in the 2D projections.

Automatic extraction of 3D dynamic left ventricle model from 2D rotational angiocardiogram.

In this paper, we propose an automatic method to directly extract 3D dynamic left ventricle (LV) model from sparse 2D rotational angiocardiogram (each cardiac phase contains only five projections). The extracted dynamic model provides quantitative cardiac function for analysis. The overlay of the model onto 2D real-time fluoroscopic images provides valuable visual guidance during cardiac intervention. Though containing severe cardiac motion artifacts, an ungated CT reconstruction is used in our approach to extract a rough static LV model. The initialized LV model is projected onto each 2D projection image. The silhouette of the projected mesh is deformed to match the boundary of LV blood pool. The deformation vectors of the silhouette are back-projected to 3D space and used as anchor points for thin plate spline (TPS) interpolation of other mesh points. The proposed method is validated on 12 synthesized datasets. The extracted 3D LV meshes match the ground truth quite well with a mean point-to-mesh error of 0.51 +/- 0.11 mm. The preliminary experiments on two real datasets (included a patient and a pig) show promising results too.

Cardiac C-arm CT: a unified framework for motion estimation and dynamic CT.

Generating 3-D images of the heart during interventional procedures is a significant challenge. In addition to real time fluoroscopy, angiographic C-arm systems can also now be used to generate 3-D/4-D CT images on the same system. One protocol for cardiac CT uses ECG triggered multisweep scans. A 3-D volume of the heart at a particular cardiac phase is then reconstructed by applying Feldkamp (FDK) reconstruction to the projection images with retrospective ECG gating. In this work we introduce a unified framework for heart motion estimation and dynamic cone-beam reconstruction using motion corrections. The benefits of motion correction are 1) increased temporal and spatial resolution by removing cardiac motion which may still exist in the ECG gated data sets and 2) increased signal-to-noise ratio (SNR) by using more projection data than is used in standard ECG gated methods. Three signal-enhanced reconstruction methods are introduced that make use of all of the acquired projection data to generate a 3-D reconstruction of the desired cardiac phase. The first averages all motion corrected back-projections; the second and third perform a weighted averaging according to 1) intensity variations and 2) temporal distance relative to a time resolved and motion corrected reference FDK reconstruction. In a comparison study seven methods are compared: nongated FDK, ECG-gated FDK, ECG-gated, and motion corrected FDK, the three signal-enhanced approaches, and temporally aligned and averaged ECG-gated FDK reconstructions. The quality measures used for comparison are spatial resolution and SNR. Evaluation is performed using phantom data and animal models. We show that data driven and subject-specific motion estimation combined with motion correction can decrease motion-related blurring substantially. Furthermore, SNR can be increased by up to 70% while maintaining spatial resolution at the same level as is provided by the ECG-gated FDK. The presented framework provides excellent image quality for cardiac C-arm CT.

A deformation tracking approach to 4D coronary artery tree reconstruction.

This paper addresses reconstruction of a temporally deforming 3D coronary vessel tree, i.e., 4D reconstruction from a sequence of angiographic X-ray images acquired by a rotating C-arm. Our algorithm starts from a 3D coronary tree that was reconstructed from images of one cardiac phase. Driven by gradient vector flow (GVF) fields, the method then estimates deformation such that projections of deformed models align with X-ray images of corresponding cardiac phases. To allow robust tracking of the coronary tree, the deformation estimation is regularized by smoothness and cyclic deformation constraints. Extensive qualitative and quantitative tests on clinical data sets suggest that our algorithm reconstructs accurate 4D coronary trees and regularized estimation significantly improves robustness. Our experiments also suggest that a hierarchy of deformation models with increasing complexities are desirable when input data are noisy or when the quality of the 3D model is low.