Search

Background The interpretation of coronary computed tomography angiography (CCTA) stenosis may be difficult among radiologists of different experience levels. Artificial intelligence (AI) may improve the diagnostic performance. Purpose To investigate whether the diagnostic performance and time efficiency of radiologists with different levels of experience in interpreting CCTA images could be improved by using CCTA with AI assistance (CCTA-AI). Material and Methods This analysis included 200 patients with complete CCTA and invasive coronary angiography (ICA) data, using ICA results as the reference. Eighteen radiologists were divided into three levels based on experience (Levels I, II, and III), and the three levels were divided into groups without (Groups 1, 2, and 3) and with (Groups 4, 5, and 6) AI assistance, totaling six groups (to avoid reader recall bias). The average sensitivity, specificity, NPV, PPV, and AUC were reported for the six groups and CCTA-AI at the patient, vessel, and segment levels. The interpretation time in the groups with and without CCTA-AI was recorded. Results Compared to the corresponding group without CCTA-AI, the Level I group with CCTA-AI had improved sensitivity (75.0% vs. 83.0% on patient-based; P = 0.003). At Level III, the specificity was better with CCTA-AI. The median interpretation times for the groups with and without CCTA-AI were 413 and 615 s, respectively ( P Conclusion CCTA-AI could assist with and improve the diagnostic performance of radiologists with different experience levels, with Level I radiologists exhibiting improved sensitivity and Level III radiologists exhibiting improved specificity. The use of CCTA-AI could shorten the training time for radiologists.

Illumination normalization for face recognition is very important when a face is captured under harsh lighting conditions. Instead of designing hand-crafted features, in this paper we formulate face illumination normalization as an image-to-image translation task. A great challenge of face normalization is that human facial structures are particularly sensitive to image structure distortion, which frequently occurs in traditional image-to-image translation tasks. Unfortunately, sometimes even slight facial structure distortions may prohibit human eyes and machine face recognition methods from identifying face identities. To address this issue, a novel GAN-based network architecture called the asymmetric joint generative adversarial network (AJGAN) is developed to normalize face images under arbitrary illumination conditions, without known face geometry and albedo information. In addition, an illumination normalization GAN $G_1$ and an asymmetric relighting GAN $G_2$ that maps a frontal-illuminated image to images with various lighting conditions are incorporated in AJGAN to maintain personalized facial structures. To avoid image blurring caused by the under-constrained relighting mapping, we introduce a scheme of one-hot lighting labels into $G_2$ and enforce label classification loss. Furthermore, the number of training images starting from a very limited number of labels is dynamically extended by the combination of different lighting labels. Qualitative and quantitative experiments on three databases validate that AJGAN significantly outperforms the state-of-the-art methods.

Illumination normalization for face recognition is of great importance when the face image is taken under a harsh lighting condition and only one single image is available. In this paper, an illumination normalization method that can preserve face identity and make the processed image have photorealistic face texture is proposed. First, an illumination regression filter is proposed to remove a large number of illumination components of the input image. Then, the accelerated proximal gradient algorithm is used for low-rank decomposition of the filtered image, which further reduces the residual illumination and noise of the face image. However, after these two operations, the face image has a peeled-off appearance. To avoid the unrealistic appearance, a deep image prior synthesis process is developed to synthesize the photorealistic face texture. To accomplish this task, the filtered image, the low-rank decomposed image, the original image and any a uniformly illuminated frontal face image as reference are used as synthetic ingredients. These ingredients, as prior knowledge, synthesize a face image that has the texture features of the reference image and retains the identity of the input image. We try to bridge the appearance gap between the synthetic ingredients and the final synthetic image by a deep neural network. Instead of training the network on a large dataset, the final synthetic image is generated by iterating the parameters of the generator-structured network according to a certain proportion of the synthetic ingredients. Systematic evaluations conducted on public databases demonstrate that the proposed method is robust to illumination with better performance than state-of-the-art methods.

Background: To investigate the influence of artificial intelligent (AI) based on deep learning on the diagnostic performance and consistency of inexperienced cardiovascular radiologists.Methods: We enrolled 196 patents who had undergone both CCTA and invasive coronary angiography (ICA) within 6 months. Four readers with less cardiovascular experience (Reader 1 to Reader 4) and two cardiovascular radiologists (level II, Reader 5 and Reader 6) evaluated all images for ≥50% coronary artery stenosis, with ICA as the gold standard. Reader 3 and Reader 4 interpreted with aid from an AI system, and the other four readers interpreted without the AI system. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy (area under the receiver operating characteristic curve (AUC)) of six readers were calculated at the patient and vessel levels. Additionally, we evaluated interobserver consistency between Reader 1 and Reader 2, Reader 3 and Reader 4, and Reader 5 and Reader 6.Results: The AI system had 94% and 78% sensitivity at the patient and vessel levels, respectively, which were higher than Reader 5 and Reader 6. Reader 3 and Reader 4 aided by AI had a higher sensitivity (range: +7.2%~ +16.6% and +5.9%~ +16.1%, respectively) and NPV (range: +3.7%~ +13.4% and +2.7%~ +4.2%, respectively) than Reader 1 and Reader 2 without AI. There was good interobserver consistency between Reader 3 and Reader 4 in interpreting ≥50% stenosis (Kappa value= 0.75 and 0.80 at the patient and vessel levels, respectively). Only Reader 1 and Reader 2 had poor consistency (Kappa value= 0.25 and 0.37). Reader 5 and Reader 6 had moderate agreement (Kappa value= 0.55 and 0.61).Conclusions: Our study showed that using AI could effectively increase the sensitivity of inexperienced readers and significantly improve consistency in diagnosing coronary stenosis via CCTA.Trial registration: The clinical trial registration number: ChiCTR1900021867Name of registry: Diagnostic performance of artificial intelligence assisted coronary computed tomography angiography for the assessment of coronary atherosclerotic stenosis

Background To investigate the influence of artificial intelligence (AI) based on deep learning on the diagnostic performance and consistency of inexperienced cardiovascular radiologists. Methods We enrolled 196 patents who had undergone both coronary computed tomography angiography (CCTA) and invasive coronary angiography (ICA) within 6 months. Four readers with less cardiovascular experience (Reader 1–Reader 4) and two cardiovascular radiologists (level II, Reader 5 and Reader 6) evaluated all images for ≥ 50% coronary artery stenosis, with ICA as the gold standard. Reader 3 and Reader 4 interpreted with AI system assistance, and the other four readers interpreted without the AI system. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy (area under the receiver operating characteristic curve (AUC)) of the six readers were calculated at the patient and vessel levels. Additionally, we evaluated the interobserver consistency between Reader 1 and Reader 2, Reader 3 and Reader 4, and Reader 5 and Reader 6. Results The AI system had 94% and 78% sensitivity at the patient and vessel levels, respectively, which were higher than that of Reader 5 and Reader 6. AI-assisted Reader 3 and Reader 4 had higher sensitivity (range + 7.2–+ 16.6% and + 5.9–+ 16.1%, respectively) and NPVs (range + 3.7–+ 13.4% and + 2.7–+ 4.2%, respectively) than Reader 1 and Reader 2 without AI. Good interobserver consistency was found between Reader 3 and Reader 4 in interpreting ≥ 50% stenosis (Kappa value = 0.75 and 0.80 at the patient and vessel levels, respectively). Only Reader 1 and Reader 2 showed poor interobserver consistency (Kappa value = 0.25 and 0.37). Reader 5 and Reader 6 showed moderate agreement (Kappa value = 0.55 and 0.61). Conclusions Our study showed that using AI could effectively increase the sensitivity of inexperienced readers and significantly improve the consistency of coronary stenosis diagnosis via CCTA. Trial registration Clinical trial registration number: ChiCTR1900021867. Name of registry: Diagnostic performance of artificial intelligence-assisted coronary computed tomography angiography for the assessment of coronary atherosclerotic stenosis.

Background: We pioneered a Deep Learning-based Coronary computed tomography–angiography Automatic Reconstruction and Diagnosis System (D-CARDS) for optimizing the workflow of diagnosis in coronary artery disease. Its efficiency and performance in clinical settings were verified. Method: In the model establishment stage, D-CARDS was trained with coronary computed tomography–angiography (CCTA) imaging data from 10,410 patients divided into training, tuning, and external validation test sets in a ratio of 7:2:1. A total of 685 patients were included in the study of clinical validation. 350 CCTA cases were collected for comparison of time efficiency. Another 335 CCTA cases were selected to reveal the diagnostic performance of D-CARDS with paired invasive coronary angiography (ICA) as the reference standard. Stenosis of 50% was considered to be obstructive and 70% or more to be significantly obstructive. The diagnostic performance of D-CARDS was evaluated as the receiver operating characteristic curve (ROC) and the corresponding area under the curve (AUC) at patient, vascular and segmental bases compared to both unilateral expert and arbitrated expert results. Findings: The average time taken of CCTA procedure after scanning (post-processing and diagnostic reporting) by D-CARDS was decreased by 73.3% from an average of 16.1 min with the conventional approach to 4.3 min (p=0.000). A total of 335 patients with 1,222 vessels and 3,559 segments were included in the final comparison of diagnostic performance. D-CARDS showed greater sensitivity (89.3% and 72.4%) than arbitrated expert results(82.6% and 62.0%)for detecting stenosis at both the 50% and 70% thresholds on patient-base, whereas its specificities were lower on every bases. The AUC showed that its diagnostic performance was equivalent to that of either of a unilateral expert on patient-based analysis, although slightly inferior to the arbitrated expert results. Interpretation: D-CARDS greatly improves the efficiency of CCTA procedure. Its diagnostic performance in detecting coronary stenosis is closer to that of an attending radiologist on patient-based analysis. The system can be used to optimize CCTA workflow. Funding Statement: This study was funded by Beijing science and technology committee. (grant reference number, Z201100005620009). Declaration of Interests: The authors declare no competing interests. Ethics Approval Statement: The study was approved by the Institutional Review Board (IRB)/Ethics Committee. The work was conducted in a manner compliant with the Measures for the Ethical Review of Biomedical Research Involving Humans and was adherent to the tenets of the Declaration of Helsinki.

Objectives To develop a computer-aided diagnosis (CAD) system for distinguishing malignant from benign pulmonary nodules on computed tomography (CT) scans, and to assess whether the diagnostic performance of radiologists with different experiences can be improved with the assistant of CAD. Materials and methods A total of 857 malignant nodules from 601 patients and 426 benign nodules from 278 patients were retrospectively collected from four hospitals. In this study, we exploited convolutional neural network in the framework of deep learning to classify whether a nodule was benign or malignant. A total of 745 malignant nodules and 370 benign nodules were used as the training data of our CAD system. The remaining 112 malignant nodules and 56 benign nodules were used as the test data. The participants were two senior chest radiologists, two secondary chest radiologists, and two junior radiology residents. The readers estimated the likelihood of malignancy of pulmonary nodules first without and then with CAD output. Receiver-operating characteristic (ROC) curve was used to evaluate readers' diagnostic performance. Results When a threshold level of 58% was used to estimate the likelihood of malignancy, the sensitivity, specificity, and diagnostic accuracy values of our CAD scheme alone were 93.8%, 83.9%, and 90.5%, respectively. For all six readers, the mean area under the ROC curve (Az ) values without and with CAD system were 0.913 and 0.938, respectively. For each reader, there is a large difference in Az values that assessed without and with CAD system. With CAD output, the readers made correct changes an average of 15.7 times and incorrect changes an average of 2 times. Conclusion Our CAD system significantly improved the diagnostic performance of readers regardless of their experience levels for assessment of the likelihood of malignancy of pulmonary nodules.

Background: Experimental animal model studies have shown neuroprotective properties of magnesium. We assessed the relationship between admission magnesium and admission stroke severity and 3-month clinical outcomes in patients with acute intracerebral hemorrhage (ICH). Methods: The present study included 323 patients with acute ICH who were prospectively identified. Demographic characteristics, lifestyle risk factors, National Institute of Health Stroke Scale (NIHSS) score, hematoma volumes, and other clinical features were recorded at baseline for all participants. Patients were divided into three groups based on the admission magnesium levels (T1: Results: After 3-month follow-up, 40 (12.4%) all-cause mortality and 132 (40.9%) poor functional outcome were documented. Median NIHSS scores for each tertile (T1 to T3) were 8.0, 5.5, and 6.0, and median hematoma volumes were 10.0, 8.05, and 12.4 ml, respectively. There was no significant association between baseline NIHSS scores (P=0.176) and hematoma volumes (P=0.442) in T3 and T1 in multivariable linear regression models. Compared with the patients in T1, those in T3 were associated with less frequency of all-cause mortality [adjusted odds ratio (OR), 0.10; 95% confidence interval (CI), 0.02-0.54; P-trend=0.010] but not poor functional outcome (adjusted OR, 1.80; 95%CI, 0.71-4.56; P-trend=0.227) after adjustment for potential confounders. Conclusion: Elevated admission serum magnesium level is associated with lower odds of mortality but not poor functional outcome at 3 months in patients with acute ICH.

Deep cerebral veins have been recently associated with the severity of hemodynamic impairment in moyamoya disease. The aim of the current study was to determine the correlation of deep medullary veins (DMVs) in susceptibility-weighted imaging (SWI) with ipsilateral cerebrovascular reactivity (CVR) of and anterior cecebrocervical artery stenosis in patients with ischemic stroke. Patients with unilateral TIA or infarction who underwent 3.0 T magnetic resonance imaging SWI, digital subtraction angiography and transcranial Doppler with CO2 stimulation within the first 7 days of hospitalization were retrospectively selected. CVR and stenosis of anterior cerebrocervical arteries were compared between different DMVs stages in symptomatic hemispheres (SHs) and asymptomatic hemispheres (AHs). A total of 61 patients were subsequently included in the present study. A univariate analysis was conducted and results for age (PAHs=0.004, PSHs=0.006), hypertension (PAHs=0.008, PSHs=0.020), current smoking (PAHs=0.006, PSHs=0.021), CVR (PAHs=0.000, PSHs=0.000), and artery stenosis (PAHs=0.000, PSHs=0.000) were obtained. The results suggested statistically significant differences between DMVs grades in SHs and AHs. A subsequent multivariate analysis revealed that CVR (ORAHs=0.925, 95% CIAHs: 0.873-0.981; ORSHs=0.945, 95% CISHs: 0.896-0.996), and artery stenosis (ORAH=3.147, 95% CIAH: 1.010-9.806; ORSHs=2.882, 95% CISHs: 1.017-8.166) were independent risk factors of DMVs. In conclusion, 3.0 T SWI was useful in detecting the DMVs around the lateral ventricle in patients with atherosclerotic ischemic stroke. CVR and stenosis of anterior cerebrocervical arteries were independent risk factors for ipsilateral DMVs in SHs and AHs.

It is known that weak texture region have become the major barriers to the development of stereo matching. The lack of appropriate feature will make it difficult to find another corresponding pixel also have no feature. The fractional differential-based approach for image filtering have the capability of nonlinearly enhancing complex texture details obvious better than by traditional integral-based algorithms. In this article, the cost aggregation consists of two pieces: the weighted guided image filtering for color-scale; the image after fractional differential filtering as the guidance image used to guided image filtering (GIF) for grayscale. The aggregated values of two scales will represent the edge and weak texture area, respectively. Finally, a disparity refinement measure based on fast weighted median filtering is applied in this paper too. Performance evaluation on Middlebury data sets shows that the proposed algorithm can obtain high-quality, especially in weak texture region. It's an attractive stereo matching solution in practice for both speed and accuracy.

vt t − αvxxtt − vxx = g(v)− αg(v)xx , x ∈ R, t > 0 (1) has a unique global generalized solutions in C2([0,∞); Hs(R)) ( s > 2 is a real number ) and a unique global classical solution in C2([0,∞); Hs(R)) ( s > 2 ) . We also prove that the Cauchy problem for the equation (1) admits a unique global generalized solution in C2([0,∞); W m,p(R) ∩ L∞(R))(m > 0 is an integer,1 6 p 6 ∞) and a unique global classical solution in C2([0,∞); W m,p(R) ∩ L∞(R)) ( m > 2 + p ) .

In this paper we study the equation of viscoelasticity u t t − u x x t − F ( u x ) x = f ( x , t ) with the nonlinear damping term on the boundary. Making use of the so-called energy perturbation method and a comparison inequality we prove the relationship between decay rate of solutions and f ( x , t ) , that is, they have the same exponential decay or algebraic decay for the case F ( s ) = s + | s | α − 2 ( α ⩾ 2 ) , and for the case F ( s ) = | s | α − 2 ( α > 2 ) , they have the same algebraic decay.