Showing papers by "Qi Zhang published in 2021"

Potential application of dynamic contrast enhanced ultrasound in predicting microvascular invasion of hepatocellular carcinoma.

Abstract: OBJECTIVE To investigate the clinical value of dynamic contrast enhanced ultrasound (D-CEUS) in predicting the microvascular invasion (MVI) of hepatocellular carcinoma (HCC). PATIENTS AND METHODS In this retrospective study, 16 patients with surgery and histopathologically proved HCC lesions were included. Patients were classified according to the presence of MVI: MVI positive group (n = 6) and MVI negative group (n = 10). Contrast enhanced ultrasound (CEUS) examinations were performed within a week before surgery. Dynamic analysis was performed by VueBox® software (Bracco, Italy). Three regions of interests (ROIs) were set in the center of HCC lesions, at the margin of HCC lesions and in the surrounding liver parenchyma accordingly. Time intensity curves (TICs) were generated and quantitative perfusion parameters including WiR (wash-in rate), WoR (wash-out rate), WiAUC (wash-in area under the curve), WoAUC (wash-out area under the curve) and WiPi (wash-in perfusion index) were obtained and analyzed. RESULTS All of HCC lesions showed arterial hyperenhancement (100 %) and at the late phase as hypoenhancement (75%) in CEUS. Among all CEUS quantitative parameters, the WiAUC and WoAUC were higher in MVI positive group than in MVI negative group in the center HCC lesions (P < 0.05), WiAUC, WoAUC and WiPI were higher in MVI positive group than in MVI negative group at the margin of HCC lesions. WiR and WoR were significant higher in MVI positive group. CONCLUSIONS D-CEUS with quantitative perfusion analysis has potential clinical value in predicting the existence of MVI in HCC lesions.

Imaging Features of Fibrolamellar Hepatocellular Carcinoma with Contrast-Enhanced Ultrasound

Abstract: Purpose Fibrolamellar hepatocellular carcinoma (f-HCC) is a rare primary liver tumor. Imaging plays an important role in diagnosis. The aim of this retrospective study was to analyze contrast-enhanced ultrasound (CEUS) features of histologically proven f-HCC in comparison to benign focal nodular hyperplasia (FNH). Materials & Methods 16 patients with histologically proven f-HCC lesions and 30 patients with FNH lesions were retrospectively reviewed regarding CEUS features to determine the malignant or benign nature of the focal liver lesions (FLL). Five radiologists assessed the CEUS enhancement pattern and came to a consensus using the EFSUMB (European Federation of Societies for Ultrasound in Medicine and Biology) guideline criteria. Results Fibrolamellar hepatocellular carcinoma manifested as a single and huge FLL. On CEUS, f-HCC showed heterogeneous hyperenhancement in the arterial phase and hypoenhancement (16/16, 100 %) in the portal venous and late phases (PVLP) as a sign of malignancy. In contrast to the hypoenhancement of f-HCC in the PVLP, all patients with FNH showed hyperenhancement as the most distinctive feature (P 0.05). Conclusion By analyzing the hypoenhancement in the PVLP, CEUS imaging reliably diagnosed f-HCC as a malignant FLL. CEUS also showed differentiation between f-HCC and FNH lesions, showing similar non-enhanced central scars, whereas f-HCC lesions showed peripheral hyperenhancement in the arterial phase and early washout in the PVLP.

Development of deep learning algorithms for predicting blastocyst formation and quality by time-lapse monitoring.

Abstract: Approaches to reliably predict the developmental potential of embryos and select suitable embryos for blastocyst culture are needed. The development of time-lapse monitoring (TLM) and artificial intelligence (AI) may help solve this problem. Here, we report deep learning models that can accurately predict blastocyst formation and usable blastocysts using TLM videos of the embryo's first three days. The DenseNet201 network, focal loss, long short-term memory (LSTM) network and gradient boosting classifier were mainly employed, and video preparation algorithms, spatial stream and temporal stream models were developed into ensemble prediction models called STEM and STEM+. STEM exhibited 78.2% accuracy and 0.82 AUC in predicting blastocyst formation, and STEM+ achieved 71.9% accuracy and 0.79 AUC in predicting usable blastocysts. We believe the models are beneficial for blastocyst formation prediction and embryo selection in clinical practice, and our modeling methods will provide valuable information for analyzing medical videos with continuous appearance variation.

Artificial intelligence-based myocardial texture analysis in etiological differentiation of left ventricular hypertrophy.

Abstract: Background Transthoracic echocardiography (TTE) is widely used in clinics to evaluate left ventricular hypertrophy (LVH). However, TTE is usually insufficient for the etiological diagnoses when morphological and functional features are nonspecific. With the booming of computer science and artificial intelligence (AI), previous literature has reported the application of radiomics based on cardiac magnetic resonance imaging, cardiac computed tomography and TTE in diagnosing several myocardial abnormalities, such as myocardial infarction, myocarditis, cardiac amyloidosis, and hypertrophic cardiomyopathy (HCM). In this study, we explored the possibility of using myocardial texture features in differentiating HCM, hypertensive heart disease (HHD) and uremic cardiomyopathy (UCM) based on echocardiography. To our knowledge, this was the first study to explore TTE myocardial texture analysis for multiple LVH etiology differentiation. Methods TTE images were reviewed retrospectively from January 2018 to collect 50 cases for each group of HHD, HCM and UCM. The apical four chamber view was retrieved. Seventeen first-order statistics and 60 gray level co-occurrence matrix (GLCM) features were extracted for statistics and classification test by support vector machine (SVM). Results Of all the parameters, entropy of brightness (EtBrt), standard deviation (Std), coefficient of variation (CoV), skewness (Skew), contrast7 (Cont7) and homogeneity5 (Hm5) were found statistically significant among the three groups (all P 0.50). As a result, HCM showed the most homogeneous myocardial texture, and was significantly different from HHD and UCM (all six features: P≤0.005). HHD appeared slightly more homogeneous than UCM, as only EtBrt and CoV were significant (P=0.011 and P=0.008). According to higher areas under the receiver operating characteristic curve (AUC) (>0.50), EtBrt, Std, and CoV were selected for test of classification as a combination of features. The AUC derived from SVM model was slightly improved compared with those of EtBrt, Std and CoV individually. Conclusions AI-based myocardial texture analysis using ultrasonic images may be a potential approach to aiding LVH etiology differentiation.

High-Frame Rate Vector Flow Imaging Technique: Initial Application in Evaluating the Hemodynamic Changes of Carotid Stenosis Caused by Atherosclerosis

Abstract: Objective: To investigate the value of high-frame rate vector flow imaging technique (V flow) in evaluating the hemodynamic changes of carotid stenosis caused by atherosclerotic plaques. Methods and Materials: In this prospective study, patients with stenosis rate (diameter) ≥30% caused by carotid atherosclerotic plaques were included. Degrees of carotid stenosis were graded according to North American Symptomatic Carotid Endarterectomy Trial criteria: moderate (30-69%) or severe (70-99%). Mindray Resona 7s ultrasound machine with a linear array transducer (3-11 MHz) was used for ultrasound examinations. The mean WSS value of carotid arteries was measured at the proximal, narrowest region and distal of carotid stenosis. The mean WSS values were correlated with peak systolic velocity (PSV) measured by color Doppler flow imaging and stenosis degree detected by digital subtraction angiography (DSA). The vector arrows and flow streamline detected by V flow dynamic imaging were analyzed. Imaging findings of DSA in carotid arteries were used as the gold standard. Results: Finally, 51 patients were included. V flow measurements were performed successfully in 17 patients (100%) with moderate-grade stenosis and in 30 patients (88.2%) with severe-grade stenosis. Dynamic V flow imaging showed yellow or red vectors at the stenotic segment, indicating fast speed blood flow (up to 260.92 cm/s). Changes of streamlines were detected in the stenotic segment. The mean WSS value measured at the narrowest region of the carotid artery had a moderately positive correlation with stenosis degree (r = 0.58, P < 0.05) and PSV value (r = 0.54, P < 0.05), respectively. Significant difference was detected in mean WSS value at the narrowest region of the carotid artery between severe carotid stenosis (1.47 ± 0.97 Pa) and moderate carotid stenosis (0.96 ± 0.44 Pa) (P < 0.05). Conclusion: The hemodynamic changes detected by V flow of the carotid stenosis might be a potential non-invasive imaging tool for assessing the degree of carotid stenosis.

Effects of Intracoronary Pro-urokinase or Tirofiban on Coronary Flow During Primary Percutaneous Coronary Intervention for Acute Myocardial Infarction: A Multi-Center, Placebo-Controlled, Single-Blind, Randomized Clinical Trial

Abstract: Background: To determine whether intracoronary pro-urokinase or tirofiban improves myocardial reperfusion during primary percutaneous coronary intervention (PCI) for acute ST-segment elevation myocardial infarction (STEMI). Methods: The study included patients with acute STEMI presenting within 12 h of symptoms at 11 hospitals in China between November 2015 and July 2017. Patients were randomized to receive selective intracoronary infusion of recombinant pro-urokinase (20 mg), tirofiban (10 μg/kg), or saline (20 mL) proximal to the infarct-related lesion over a 3-min period before stent implantation during primary PCI. The primary outcome was final corrected thrombolysis in myocardial infarction (TIMI) frame count (CTFC) after PCI. Results: This study included 345 patients. Initial angiography identified a high-grade thrombus (TIMI 4-5) in 80% of patients. Final CTFC after PCI was significantly lower in the pro-urokinase (P 0.05). The pro-urokinase (P = 0.008) and tirofiban groups (P = 0.022) had more complete ST-segment resolution at 2 h and lower peak creatine kinase-MB levels after PCI than the saline group (P = 0.006 and P = 0.023). The 30-day incidence of major adverse cardiac events was 4.5% in the pro-urokinase group, 3.4% in the tirofiban group, and 2.6% in the saline group. The incidence of in-hospital TIMI major bleeding events was low and comparable between groups. Conclusions: Adjunctive intracoronary pro-urokinase or tirofiban given before stent implantation during primary PCI improves myocardial reperfusion without increasing the incidence of major bleeding events.

The value of dynamic contrast enhanced ultrasound (DCE-US) in monitoring treatment effect of high-intensity focused ultrasound (HIFU) in locally advanced pancreatic cancer (LAPC).

Abstract: Purpose To evaluate the feasibility of dynamic contrast enhanced ultrasound (DCE-US) in predicting treatment response of high-intensity focused ultrasound (HIFU) in patients with locally advanced pancreatic cancer (LAPC) lesions. Patients and methods In this prospective study, 10 patients with pathologically confirmed LAPC lesions (7 men, 3 women; average age, 61.13±5.80 years) were prospectively enrolled. All patients received HIFU treatment with peak intensity at 12000 W/cm2. Contrast enhanced ultrasound (CEUS) was performed with an ACUSON Oxana 2 ultrasound equipment and a 6 C-1 transducer (1-6 Hz). A dose of 2.4 ml SonoVue was injected for each examination. Time intensity curves (TICs) were generated and quantitative analyses were performed by SonoLiver software. B mode ultrasound (BMUS) features, CEUS enhancement patterns, TICs, CEUS quantitative parameters and serum carcinoma antigen 19-9 (CA19-9) levels were compared before and 4 weeks after HIFU treatment. Statistical analyses were performed with SPSS Version 20.0 and GraphPad Prism 5. Results While comparing before and after HIFU, no significant difference was obtained on mean size of lesion, BMUS or CEUS features. After HIFU treatment, TICs showed decreased and delayed enhancement. Among all CEUS quantitative parameters, significant decrease could be found in maximum intensity (MI) (60.66±23.95% vs 41.31±26.74%) and mean transit time (mTT) (76.66±47.61 s vs 38.42±28.35 s). CA19-9 level decreased significantly after HIFU (2747.92±4237.41 U/ml vs 715.08±1773.90 U/ml) (P = 0.05). Conclusion DCE-US combining with quantitative analysis might be a useful imaging method for early treatment response evaluation of HIFU in LAPC lesions.

Customizing Robot-Assisted Passive Neurorehabilitation Exercise Based on Teaching Training Mechanism.

Abstract: Passive movement is an important mean of rehabilitation for stroke survivors in the early stage or with greater paralysis. The upper extremity robot is required to assist therapists with passive movement during clinical rehabilitation, while customizing is one of the crucial issues for robot-assisted upper extremity training, which fits the patient-centeredness. Robot-assisted teaching training could address the need well. However, the existing control strategies of teaching training are usually commanded by position merely, having trouble to achieve the efficacy of treatment by therapists. And deficiency of flexibility and compliance comes to the training trajectory. This research presents a novel motion control strategy for customized robot-assisted passive neurorehabilitation. The teaching training mechanism is developed to coordinate the movement of the shoulder and elbow, ensuring the training trajectory correspondence with human kinematics. Furthermore, the motion trajectory is adjusted by arm strength to realize dexterity and flexibility. Meanwhile, the torque sensor employed in the human-robot interactive system identifies movement intention of human. The goal-directed games and feedbacks promote the motor positivity of stroke survivors. In addition, functional experiments and clinical experiments are investigated with a healthy adult and five recruited stroke survivors, respectively. The experimental results present that the suggested control strategy not only serves with safety training but also presents rehabilitation efficacy.

3-D Gabor-based anisotropic diffusion for speckle noise suppression in dynamic ultrasound images.

Abstract: Speckle noise contaminates medical ultrasound images, and the suppression of speckle noise is helpful for image interpretation. Traditional ultrasound denoising (i.e., despeckling) methods are developed on two-dimensional static images. However, one of the advantages of ultrasonography is its nature of dynamic imaging. A method for dynamic ultrasound despeckling is expected to incorporate both the spatial and temporal information in successive images of dynamic ultrasound and thus yield better denoising performance. Here we regard a dynamic ultrasound video as three-dimensional (3-D) images with two dimensions in the spatial domain and one in the temporal domain, and we propose a despeckling algorithm for dynamic ultrasound named the 3-D Gabor-based anisotropic diffusion (GAD-3D). The GAD-3D expands the classic two-dimensional Gabor-based anisotropic diffusion (GAD) into 3-D domain. First, we proposed a robust 3-D Gabor-based edge detector by capturing the edge with 3-D Gabor transformation. Then we embed this novel detector into the partial differential equation of GAD to guide the 3-D diffusion process. In the simulation experiment, when the noise variance is as high as 0.14, the GAD-3D improves the Pratt's figure of merit, mean structural similarity index and peak signal-to-noise ratio by 24.32%, 10.98%, and 6.51%, respectively, compared with the best values of seven other methods. Experimental results on clinical dynamic ultrasonography suggest that the GAD-3D outperforms the other seven methods in noise reduction and detail preservation. The GAD-3D is effective for dynamic ultrasound despeckling and may be potentially valuable for disease assessment in dynamic medical ultrasonography.

Segmentation of Lymph Nodes in Ultrasound Images Using U-Net Convolutional Neural Networks and Gabor-Based Anisotropic Diffusion

Abstract: The automated segmentation of lymph nodes (LNs) in ultrasound images is challenging, largely because of speckle noise and echogenic hila. This paper proposes a fully automatic and accurate method for LN segmentation in ultrasound that overcomes these issues. The proposed segmentation method integrates diffusion-based despeckling, U-Net convolutional neural networks and morphological operations. First, the speckle noise is suppressed and the lymph node edges are enhanced using Gabor-based anisotropic diffusion (GAD). Then, a modified U-Net model is used to segment the LNs excluding any echogenic hila. Finally, morphological operations are undertaken to segment the entire LNs by filling in any regions occupied by echogenic hila. A total of 531 lymph nodes from 526 patients were segmented using the proposed method. Its segmentation performance was evaluated in terms of its accuracy, sensitivity, specificity, Jaccard similarity and Dice coefficient, for which it achieved values of 0.934, 0.939, 0.937, 0.763 and 0.865, respectively. The proposed method automatically and accurately segments LNs in ultrasound images, enhancing the prospects of being able to undertake artificial intelligence (AI)-based diagnosis of lymph node diseases.

Diabetic Peripheral Neuropathy Affects Pinch Strength and Hand Dexterity in Elderly Patients.

Abstract: Objective Diabetic peripheral neuropathy (DPN) is one of the most common chronic complications of diabetes, leading to disability and decreased quality of life. In past research and clinical studies, the lower limb function of DPN patients was often the principal subject of research, with little attention given to the upper limb and hand. Our goal was to assess and compare hand function between elderly diabetic patients with DPN and without DPN. Methods A total of 52 diabetic patients were registered and underwent hand function assessments and electrodiagnostic tests. Dynamometer, pinch meter, Semmes Weinstein monofilaments, and the Purdue Pegboard Test (PPT) were used to assess the patients' grip strength, pinch strength, tactile sensory threshold, and hand dexterity. Results Compared with the non-DPN group, the elderly DPN group showed worse thumb-middle fingertip pinch strength and thumb-little fingertip pinch strength in the dominant hand (3.50 (2.50, 4.25) vs. 4.50 (3.00, 5.00), p = 0.019; 1.50 (1.00, 2.00) vs. 2.50 (2.00, 3.00), p < 0.001); the elderly DPN group displayed worse thumb-middle fingertip pinch strength, thumb-ring fingertip pinch strength, and thumb-little fingertip pinch strength in the nondominant hand (3.50 (2.00, 4.50) vs. 4.00 (3.00, 5.00), p = 0.013; 2.50 (1.25, 3.00) vs. 3.00 (2.50, 3.50), p = 0.033; 1.00 (0.75, 2.25) vs. 2.50 (2.00, 2.50), p < 0.001). The elderly DPN group scored lower than the non-DPN group on the PPT test of assembly (13.96 ± 5.18 vs. 16.96 ± 4.61, t = 2.212, p = 0.032). Conclusion Motor function limitation is the principal hand dysfunction in elderly patients with DPN, which is mainly manifested as a decline in fingertip pinch strength and a decrease in hand dexterity. This trial is registered with Clinical Trial Registry no. ChiCTR1900025358.

Longitudinal trajectories of pneumonia lesions and lymphocyte counts associated with disease severity among convalescent COVID-19 patients: a group-based multi-trajectory analysis.

Abstract: To explore the long-term trajectories considering pneumonia volumes and lymphocyte counts with individual data in COVID-19. A cohort of 257 convalescent COVID-19 patients (131 male and 126 females) were included. Group-based multi-trajectory modelling was applied to identify different trajectories in terms of pneumonia lesion percentage and lymphocyte counts covering the time from onset to post-discharge follow-ups. We studied the basic characteristics and disease severity associated with the trajectories. We characterised four distinct trajectory subgroups. (1) Group 1 (13.9%), pneumonia increased until a peak lesion percentage of 1.9% (IQR 0.7–4.4) before absorption. The slightly decreased lymphocyte rapidly recovered to the top half of the normal range. (2) Group 2 (44.7%), the peak lesion percentage was 7.2% (IQR 3.2–12.7). The abnormal lymphocyte count restored to normal soon. (3) Group 3 (26.0%), the peak lesion percentage reached 14.2% (IQR 8.5–19.8). The lymphocytes continuously dropped to 0.75 × 109/L after one day post-onset before slowly recovering. (4) Group 4 (15.4%), the peak lesion percentage reached 41.4% (IQR 34.8–47.9), much higher than other groups. Lymphopenia was aggravated until the lymphocytes declined to 0.80 × 109/L on the fourth day and slowly recovered later. Patients in the higher order groups were older and more likely to have hypertension and diabetes (all P values < 0.05), and have more severe disease. Our findings provide new insights to understand the heterogeneous natural courses of COVID-19 patients and the associations of distinct trajectories with disease severity, which is essential to improve the early risk assessment, patient monitoring, and follow-up schedule.

Prediction of Progression to Severe Stroke in Initially Diagnosed Anterior Circulation Ischemic Cerebral Infarction.

Abstract: Purpose: Accurate prediction of the progression to severe stroke in initially diagnosed nonsevere patients with acute-subacute anterior circulation nonlacuna ischemic infarction (ASACNLII) is important in making clinical decision. This study aimed to apply a machine learning method to predict if the initially diagnosed nonsevere patients with ASACNLII would progress to severe stroke by using diffusion-weighted images and clinical information on admission. Methods: This retrospective study enrolled 344 patients with ASACNLII from June 2017 to August 2020 on admission, and 108 cases progressed to severe stroke during hospitalization within 3-21 days. The entire data were randomized into a training set (n = 271) and an independent test set (n = 73). A U-Net neural network was employed for automatic segmentation and volume measurement of the ischemic lesions. Predictive models were developed and used for evaluating the progression to severe stroke using different feature sets (the volume data, the clinical data, and the combination) and machine learning methods (random forest, support vector machine, and logistic regression). Results: The U-Net showed high correlation with manual segmentation in terms of Dice coefficient of 0.806 and R2 value of the volume measurements of 0.960 in the test set. The random forest classifier of the volume + clinical combination achieved the best area under the receiver operating characteristic curve of 0.8358 (95% CI 0.7321-0.9269), and the accuracy, sensitivity, and specificity were 0.7780 (0.7397-0.7945), 0.7695 (0.6102-0.9074), and 0.8686 (0.6923-1.0), respectively. The Shapley additive explanation diagram showed the volume variable as the most important predictor. Conclusion: The U-Net was fully automatic and showed a high correlation with manual segmentation. An integrated approach combining clinical variables and stroke lesion volumes that were derived from the advanced machine learning algorithms had high accuracy in predicting the progression to severe stroke in ASACNLII patients.

Cross-Tissue/Organ Transfer Learning for the Segmentation of Ultrasound Images Using Deep Residual U-Net

Abstract: Ultrasound image segmentation is a crucial step in computer-aided diagnosis. In this study, we propose a cross-tissue/organ segmentation method based on the transfer learning method and a modified deep residual U-Net model. We present a modified deep residual U-Net model by integrating a U-Net architecture with residual blocks to leverage the advantages of both components. Next, we explore a cross-tissue/organ transfer learning method for ultrasound image segmentation, which transfers the knowledge of ultrasound image segmentation from one tissue/organ to another, e.g. from tendon images to breast tumor images and vice versa. We evaluated the proposed method by performing four groups of experiments on three medical ultrasound datasets, consisting of one tendon dataset and two breast datasets, along with one non-medical dataset. The results showed an overall performance improvement by our method in terms of the Dice coefficient and Jaccard index. It was demonstrated that our modified deep residual U-Net exceeded the standard U-Net and residual U-Net, and the cross-tissue/organ transfer learning was superior to training from scratch and to transfer learning between divergent domains. Our method shows potential to accurately segment medical ultrasound images.

Clinical and Genetic Predictive Models for the Prediction of Pathological Complete Response to Optimize the Effectiveness for Trastuzumab Based Chemotherapy.

Abstract: Background Trastuzumab shows excellent benefits for HER2+ breast cancer patients, although 20% treated remain unresponsive. We conducted a retrospective cohort study to optimize neoadjuvant chemotherapy and trastuzumab treatment in HER2+ breast cancer patients. Methods Six hundred patients were analyzed to identify clinical characteristics of those not achieving a pathological complete response (pCR) to develop a clinical predictive model. Available RNA sequence data was also reviewed to develop a genetic model for pCR. Results The pCR rate was 39.8% and pCR was associated with superior disease free survival and overall survival. ER negativity and PR negativity, higher HER2 IHC scores, higher Ki-67, and trastuzumab use were associated with improved pCR. Weekly paclitaxel and carboplatin had the highest pCR rate (46.70%) and the anthracycline+taxanes regimen had the lowest rate (11.11%). Four published GEO datasets were analyzed and a 10-gene model and immune signature for pCR were developed. Non-pCR patients were ER+PR+ and had a lower immune signature and gene model score. Hormone receptor status and immune signatures were independent predictive factors of pCR. Conclusion Hormone receptor status and a 10-gene model could predict pCR independently and may be applied for patient selection and drug effectiveness optimization.