GZGuangyu Zhu

Guangyu Zhu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Guangyu Zhu has a total of 56 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Defect-Specific Physical-Prior-Guided Synthesis of Photovoltaic RGB Defect Images with Pixel-Level Annotations

    by Xujiang Liu, Tingting Yang, Guangyu Zhu

    Reliable visual inspection of photovoltaic modules is limited by the scarcity of diverse defect images and pixel-level annotations. This study proposes a defect-specific RGB image synthesis framework combining local feature disentanglement and physical-prior-guided modeling for shadow, debris, dust, and broken defects, while simultaneously generating spatially aligned pixel-level masks. The generated samples are evaluated in terms of image quality, distribution consistency, structural preservation, and downstream utility. Compared with CycleGAN, MUNIT, and SDXL Inpainting, the proposed method achieves the lowest BRISQUE scores across all four categories, the lowest class-wise Defect-aware Fréchet Distance (DFD) for shadow, debris, and broken, and the lowest MMD2-RBF values across all categories. With generated samples added to real training data, the five-run average classification Accuracy and Macro-F1 reach 88.96% and 87.83%, respectively, while detection performance is also improved. Moreover, a segmentation model trained solely on generated image–mask pairs achieves an mIoU of 60.15% on unseen real images. These results demonstrate that the proposed framework provides structurally consistent, task-useful defect samples and directly usable pixel-level supervision for photovoltaic visual inspection.

  • Open Access

    Clinical value of aortic arch morphology in transfemoral TAVR: artificial intelligence evaluation

    by Yu Mao, Yang Liu, Mengen Zhai, Ping Jin, Fangyao Chen, Yuhui Yang, Guangyu Zhu, Tingting Yang, Gejun Zhang, Kai Xu, et al.

    Background: The impact of aortic arch (AA) morphology on the management of the procedural details and the clinical outcomes of the transfemoral artery (TF)-transcatheter aortic valve replacement (TAVR) has not been evaluated. The goal of this study was to evaluate the AA morphology of patients who had TF-TAVR using an artificial intelligence algorithm and then to evaluate its predictive value for clinical outcomes. Materials and methods: A total of 1480 consecutive patients undergoing TF-TAVR using a new-generation transcatheter heart valve at 12 institutes were included in this retrospective study. The AA measurements were evaluated by deep learning, and then the approach index (IA) was determined. The machine learning algorithm was used to construct the predictive model and was validated externally. Results: The area under the curve of the IA model using random forest and logistic regression was 0.675 [95% confidence interval (CI): 0.586–0.764] and 0.757 (95% CI: 0.665–0.849), respectively. The IA model was validated externally, and consistent distinctions were obtained. After we used a generalized propensity score matching method for continuous exposure, the IA was the strongest correlation factor for major procedural events (odds ratio: 3.87; 95% CI: 2.13–7.59, P < 0.001). When leaflet morphology or transcatheter heart valve type was an interactive item with IA, neither of them was statistically significant in terms of clinical outcomes. Conclusion: IA may be used to identify the impact of AA morphology on procedural and clinical outcomes in patients having TF-TAVR and to help to predict the procedural complications.


  • Open Access

    A benchmark study of convolutional neural networks in fully automatic segmentation of aortic root

    by Tingting Yang, Guangyu Zhu, Li Cai, Joon Hock Yeo, Yu Mao, Jian Yang

    Recent clinical studies have suggested that introducing 3D patient-specific aortic root models into the pre-operative assessment procedure of transcatheter aortic valve replacement (TAVR) would reduce the incident rate of peri-operative complications. Tradition manual segmentation is labor-intensive and low-efficient, which cannot meet the clinical demands of processing large data volumes. Recent developments in machine learning provided a viable way for accurate and efficient medical image segmentation for 3D patient-specific models automatically. This study quantitively evaluated the auto segmentation quality and efficiency of the four popular segmentation-dedicated three-dimensional (3D) convolutional neural network (CNN) architectures, including 3D UNet, VNet, 3D Res-UNet and SegResNet. All the CNNs were implemented in PyTorch platform, and low-dose CTA image sets of 98 anonymized patients were retrospectively selected from the database for training and testing of the CNNs. The results showed that despite all four 3D CNNs having similar recall, Dice similarity coefficient (DSC), and Jaccard index on the segmentation of the aortic root, the Hausdorff distance (HD) of the segmentation results from 3D Res-UNet is 8.56 ± 2.28, which is only 9.8% higher than that of VNet, but 25.5% and 86.4% lower than that of 3D UNet and SegResNet, respectively. In addition, 3D Res-UNet and VNet also performed better in the 3D deviation location of interest analysis focusing on the aortic valve and the bottom of the aortic root. Although 3D Res-UNet and VNet are evenly matched in the aspect of classical segmentation quality evaluation metrics and 3D deviation location of interest analysis, 3D Res-UNet is the most efficient CNN architecture with an average segmentation time of 0.10 ± 0.04 s, which is 91.2%, 95.3% and 64.3% faster than 3D UNet, VNet and SegResNet, respectively. The results from this study suggested that 3D Res-UNet is a suitable candidate for accurate and fast automatic aortic root segmentation for pre-operative assessment of TAVR.


  • Open Access

    Deep learning-based recognition and segmentation of intracranial aneurysms under small sample size

    by Guangyu Zhu, Xueqi Luo, Tingting Yang, Li Cai, Joon Hock Yeo, Ge Yan, Jian Yang

    The manual identification and segmentation of intracranial aneurysms (IAs) involved in the 3D reconstruction procedure are labor-intensive and prone to human errors. To meet the demands for routine clinical management and large cohort studies of IAs, fast and accurate patient-specific IA reconstruction becomes a research Frontier. In this study, a deep-learning-based framework for IA identification and segmentation was developed, and the impacts of image pre-processing and convolutional neural network (CNN) architectures on the framework’s performance were investigated. Three-dimensional (3D) segmentation-dedicated architectures, including 3D UNet, VNet, and 3D Res-UNet were evaluated. The dataset used in this study included 101 sets of anonymized cranial computed tomography angiography (CTA) images with 140 IA cases. After the labeling and image pre-processing, a training set and test set containing 112 and 28 IA lesions were used to train and evaluate the convolutional neural network mentioned above. The performances of three convolutional neural networks were compared in terms of training performance, segmentation performance, and segmentation efficiency using multiple quantitative metrics. All the convolutional neural networks showed a non-zero voxel-wise recall (V-Recall) at the case level. Among them, 3D UNet exhibited a better overall segmentation performance under the relatively small sample size. The automatic segmentation results based on 3D UNet reached an average V-Recall of 0.797 ± 0.140 (3.5% and 17.3% higher than that of VNet and 3D Res-UNet), as well as an average dice similarity coefficient (DSC) of 0.818 ± 0.100, which was 4.1%, and 11.7% higher than VNet and 3D Res-UNet. Moreover, the average Hausdorff distance (HD) of the 3D UNet was 3.323 ± 3.212 voxels, which was 8.3% and 17.3% lower than that of VNet and 3D Res-UNet. The three-dimensional deviation analysis results also showed that the segmentations of 3D UNet had the smallest deviation with a max distance of +1.4760/−2.3854 mm, an average distance of 0.3480 mm, a standard deviation (STD) of 0.5978 mm, a root mean square (RMS) of 0.7269 mm. In addition, the average segmentation time (AST) of the 3D UNet was 0.053s, equal to that of 3D Res-UNet and 8.62% shorter than VNet. The results from this study suggested that the proposed deep learning framework integrated with 3D UNet can provide fast and accurate IA identification and segmentation.


  • Open Access

    Numerical Investigations on the Blade Tip Clearance Excitation Forces in an Unshrouded Turbine

    by Yang Pan, Qi Yuan, Gongge Huang, Jiawei Gu, Pu Li, Guangyu Zhu

    The purpose of this study was to investigate the characteristics of the blade tip excitation forces represented as the rotordynamic coefficients (stiffness and damping coefficients) in an unshrouded turbine using the three-dimensional computational fluid dynamic (CFD) numerical method. The blade geometrical parameters were based on a SNECMA transonic experimental rig. The simulations were performed by solving the compressible Reynolds-averaged Navier–Stokes (RANS) equations. The multi-frequency elliptical whirling orbit model and an improved mesh deformation method based on the transient analysis were utilized. The effects of operating conditions on the rotordynamic coefficients and the unsteady flow were also found. The results show that the positive direct stiffness, which confirmed the direct force contribution in the tip excitation forces and the cross-coupling stiffness, were dependent on the whirling frequencies. Damping effects were shown to be negligible. The rotational speed, inlet flow angle, eccentric ratio (ER), and mean tip clearance had impacts on the stiffness, and some effects of these variables on the rotordynamic coefficients were found to be frequency dependent. Additionally, increasing the rotor eccentricity and the mean tip clearance led to the nonuniformity of the circumferential pressure distributions.


  • Open Access

    Numerical Investigation of the Effects of Prosthetic Aortic Valve Design on Aortic Hemodynamic Characteristics

    by Guang-Yu Zhu, Hai Huang, Ya-Li Su, Joon-Hock Yeo, Xiao-Qin Shen, Cheng-Fu Yang

    The superior performance of single-point attached commissures (SPAC) molded valve design has been validated by several numerical, in vitro and in vivo animal studies. However, the impacts of the SPAC molded valve design on aortic hemodynamic environments are yet to be investigated. In this study, multiscale computational models were prepared by virtually implanting prosthetic aortic valves with SPAC tubular, SPAC molded and conventional designs into a patient-specific aorta, respectively. The impacts of the valve designs on efferent flow distribution, flow pattern and hemodynamic characteristics in the aorta were numerically investigated. The results showed that despite the overall flow phenomena being similar, the SPAC tubular valve exhibited a suboptimal performance in terms of higher spatially averaged wall shear stress (SAWSS) in ascending aorta (AAo), higher helix grade, stronger secondary flow mean secondary velocity in descending aorta, as well as more complex vortex distribution. The results from the current study extend the understanding of hemodynamic impacts of the valve designs, which would further benefit the optimization of the prosthetic aortic valve.


  • Open Access

    In vitro Assessment of the Impacts of Leaflet Design on the Hemodynamic Characteristics of ePTFE Pulmonary Prosthetic Valves

    by Guangyu Zhu, Yuan Wei, Qi Yuan, Li Cai, Masakazu Nakao, Joon Hock Yeo

    Prosthetic pulmonary valves are widely used in the management procedures of various congenital heart diseases, including the surgical pulmonary valve replacement (PVR) and right ventricular outflow tract reconstruction (RVOT). The discouraging long-term outcomes of standard prostheses, including homografts and bioprosthetic, constrained their indications. Recent developments in the expanded-polytetrafluoroethylene (ePTFE) pulmonary prosthetic valves provide promising alternatives. In this study, the hemodynamic characteristics of bileaflet and trileaflet ePTFE valve designs were experimentally evaluated. The in vitro tests were performed under the right ventricle (RV) flow conditions by using an in vitro RV circulatory system and particle image velocimetry (PIV). The leaflet kinetics, trans-valvular pressure gradients, effective orifice areas, regurgitant fractions, energy losses, velocity fields, and Reynolds shear stress (RSS) in both prostheses were evaluated. The opening of the bileaflet and trileaflet valve takes 0.060 and 0.088 s, respectively. The closing of the former takes 0.140 s, in contrast to 0.176 s of the latter. The trans-valvular pressure is 6.8 mmHg in the bileaflet valve vs. 7.9 mmHg in the trileaflet valve. The effective orifice area is 1.83 cm2 in the bileaflet valve and 1.72 cm2 in the trileaflet valve. The regurgitant fraction and energy loss of bileaflet are 7.13% and 82 mJ, which are 7.84% and 101.64 mJ in its bileaflet counterpart. The maximum RSS of 48.0 and 49.2 Pa occur at the systole peak in the bileaflet and trileaflet valve, respectively. A higher average RSS level is found in the bileaflet valve. The results from this preliminary study indicate that the current bileaflet prosthetic valve design is capable of providing a better overall hemodynamic performance than the trileaflet design.


  • Open Access

    Impacts of Internal Carotid Artery Revascularization on Flow in Anterior Communicating Artery Aneurysm: A Preliminary Multiscale Numerical Investigation

    by Guang-Yu Zhu, Yuan Wei, Ya-Li Su, Qi Yuan, Cheng-Fu Yang

    The optimal management strategy of patients with concomitant anterior communicating artery aneurysm (ACoAA) and internal carotid artery (ICA) stenosis is unclear. This study aims to evaluate the impacts of unilateral ICA revascularization on hemodynamics factors associated with rupture in an ACoAA. In the present study, a multiscale computational model of ACoAA was developed by coupling zero-dimensional (0D) models of the cerebral vascular system with a three-dimensional (3D) patient-specific ACoAA model. Distributions of flow patterns, wall shear stress (WSS), relative residence time (RRT) and oscillating shear index (OSI) in the ACoAA under left ICA revascularization procedure were quantitatively assessed by using transient computational fluid dynamics (CFD) simulations. Our results showed that the revascularization procedures significantly changed the hemodynamic environments in the ACoAA. The flow disturbance in the ACoAA was enhanced by the resumed flow from the affected side. In addition, higher OSI (0.057 vs. 0.02), prolonged RRT (1.14 vs. 0.39) and larger low WSS area (66 vs. 50 mm2) in ACoAA were found in the non-stenotic case. These acute changes in hemodynamics after revascularization may elevate the rupture risk of ACoAA. The preliminary results validated the feasibility of predicting aneurismal hemodynamics characteristics in revascularization procedures by using multiscale CFD simulations, which would benefit the management of this group of patients.


  • Open Access

    PIV investigation of the flow fields in subject-specific vertebro-basilar (VA-BA) junction

    by Guangyu Zhu, Yuan Wei, Qi Yuan, Jian Yang, Joon Hock Yeo

    As the only arterial structure of which two main arteries merged into one, the vertebro-basilar (VA-BA) system is one of the favorite sites of cerebral atherosclerotic plaques. The aim of this study was to investigate the detailed hemodynamics characteristics in the VA-BA system. A scale-up subject-specific flow phantom of VA-BA system was fabricated based on the computed tomography angiography (CTA) scanning images of a healthy adult. Flow fields in eight axial planes and six radial planes were measured and analyzed by using particle image velocimetry (PIV) under steady flow conditions of $${Re}=300$$ , $${Re}=500$$ . A water–glycerin mixture was used as the working fluid. The flow in the current model exhibited highly three-dimensional characteristics. The confluence of VAs flow formed bimodal velocity distribution near the confluence apex. Due to the asymmetrical structural configuration, the bimodal velocity profile skewed towards left, and sharper peaks were observed under higher Reynolds condition. Secondary flow characterized by two vortices formed in the radial planes where 10 mm downstream the confluence apex and persists along the BA under both Reynolds numbers. The strength of secondary flow under $${Re}=500$$ is around 8% higher than that under $${Re}=300$$ , and decayed nonlinearly along the flow direction. In addition, a low momentum recirculation region induced by boundary layer separation was observed near the confluence apex. The wall shear stress (WSS) in the recirculation area was found to be lower than 0.4 Pa. This region coincides well with the preferential site of vascular lesions in the VA-BA system. This preliminary study verified that the subject-specific in-vitro experiment is capable of reflecting the detailed flow features in the VA-BA system. The findings from this study may help to expand the understanding of the hemodynamics in the VA-BA system, and further clarifying the mechanism that underlying the localization of vascular lesions.