Search

An L1/L2-norm-bound extreme learning machine classification algorithm is proposed to improve the accuracy of distinguishing between benign and malignant pulmonary nodules. In this algorithm, features extracted from the segmented lung nodule using the histogram of oriented gradients method are used as inputs. L1-norm can promote sparsity in the weights of the output layer, and L2-norm can smooth output weights. The combination of the L1 norm and L2 norm can simplify the complexity of the network and prevent overfitting to improve classification accuracy. For each newly tested lung nodule, the algorithm outputs a class label of either benign or malignant. The accuracy, sensitivity, and specificity reached 94.12%, 93%, and 95% respectively over the lung image database consortium and image database resource initiative dataset. Compared with other algorithms, the average values of the three metrics increased by 6.5%, 7.94%, and 4.32%, respectively. An accuracy score of 95.83% can be achieved over a set of 120 urinary sediment images. Therefore, this algorithm has a good classification effect of pulmonary nodules.

Background For the encoding part of U-Net3+,the ability of brain tumor feature extraction is insufficient, as a result, the features can not be fused well during up-sampling, and the accuracy of segmentation will reduce. Methods In this study, we put forward an improved U-Net3+ segmentation network based on stage residual. In the encoder part, the encoder based on the stage residual structure is used to solve the vanishing gradient problem caused by the increasing in network depth, and enhances the feature extraction ability of the encoder which is instrumental in full feature fusion when up-sampling in the network. What’s more, we replaced batch normalization (BN) layer with filter response normalization (FRN) layer to eliminate batch size impact on the network. Based on the improved U-Net3+ two-dimensional (2D) model with stage residual, IResUnet3+ three-dimensional (3D) model is constructed. We propose appropriate methods to deal with 3D data, which achieve accurate segmentation of the 3D network. Results The experimental results showed that: the sensitivity of WT, TC, and ET increased by 1.34%, 4.6%, and 8.44%, respectively. And the Dice coefficients of ET and WT were further increased by 3.43% and 1.03%, respectively. To facilitate further research, source code can be found at: https://github.com/YuOnlyLookOne/IResUnet3Plus. Conclusion The improved network has a significant improvement in the segmentation task of the brain tumor BraTS2018 dataset, compared with the classical networks u-net, v-net, resunet and u-net3+, the proposed network has smaller parameters and significantly improved accuracy.

Medical image segmentation is a fundamental step in medical analysis and diagnosis. In recent years, deep learning networks have been used for precise segmentation. Numerous improved encoder–decoder structures have been proposed for various segmentation tasks. However, high-level features have gained more research attention than the abundant low-level features in the early stages of segmentation. Consequently, the learning of edge feature maps has been limited, which can lead to ambiguous boundaries of the predicted results. Inspired by the encoder–decoder network and attention mechanism, this study investigates a novel multilayer edge attention network (MEA-Net) to fully utilize the edge information in the encoding stages. MEA-Net comprises three major components: a feature encoder module, a feature decoder module, and an edge module. An edge feature extraction module in the edge module is designed to produce edge feature maps by a sequence of convolution operations so as to integrate the inconsistent edge information from different encoding stages. A multilayer attention guidance module is designed to use each attention feature map to filter edge information and select important and useful features. Through experiments, MEA-Net is evaluated on four medical image datasets, including tongue images, retinal vessel images, lung images, and clinical images. The evaluation values of the Accuracy of four medical image datasets are 0.9957, 0.9736, 0.9942, and 0.9993, respectively. The values of the Dice coefficient are 0.9902, 0.8377, 0.9885, and 0.9704, respectively. Experimental results demonstrate that the network being studied outperforms current state-of-the-art methods in terms of the five commonly used evaluation metrics. The proposed MEA-Net can be used for the early diagnosis of relevant diseases. In addition, clinicians can obtain more accurate clinical information from segmented medical images.

Accurate and robust segmentation of anatomical structures from magnetic resonance images is valuable in many computer-aided clinical tasks. Traditional codec networks are not satisfactory because of their low accuracy of edge segmentation, the low recognition rate of the target, and loss of detailed information. To address these problems, this study proposes a series of improved models for semantic segmentation and progressively optimizes them from the three aspects of convolution module, codec unit, and feature fusion. Instead of the standard convolution structure, we apply a new type of convolution module for the feature extraction. The networks integrate a multi-path method to obtain richer-detail edge information. Finally, a dense network is utilized to strengthen the ability of the feature fusion and integrate more different-level information. The evaluation of the Accuracy, Dice coefficient, and Jaccard index led to values of 0.9855, 0.9185, and 0.8507, respectively. These metrics of the best network increased by 1.0%, 4.0%, and 6.1%, respectively. Boundary F1-Score reached 0.9124 indicating that the proposed networks can segment smaller targets to obtain smoother edges. Our methods obtain more key information than traditional methods and achieve superiority in segmentation performance.

Highlights • There has been significant heterogeneity in the definition of what constitutes as low volume operator /center for AF ablation. • Current literature suggests lower success rates but higher complications rates in low-volume centers. • First AF catheter ablation in a low-volume center has comparable safety and efficacy outcomes to high-volume centers using contemporary ablation technologies., Background Catheter ablation is increasingly being performed worldwide for atrial fibrillation (AF). However, there are concerns of lower success rates and higher complications of AF ablations performed in low-volume centers. Thus, we sought to evaluate the safety and efficacy of AF catheter ablation in a low-volume center using contemporary technologies. Methods and results 71 consecutive patients (50 paroxysmal AF [pAF] vs 21 persistent AF) who underwent first catheter ablation were studied. Primary outcome was AF recurrence rate. Secondary outcomes included periprocedural complications, hospitalization for symptomatic tachy-arrhythmias post-ablation and number of repeat ablations. Mean age of our cohort was 59.1 ± 9.7 years, of which 56 (78.9%) were males. 1-year AF recurrence was 19.5% in pAF and 23.8% in persistent AF (p = 0.694). Ablation in persistent AF group required longer procedural (197.76 ± 48.60 min [pAF] vs 238.67 ± 70.50 min [persistent AF], p = 0.006) and ablation duration (35.08 ± 15.84 min [pAF] vs 52.65 ± 28.46 min [persistent AF], p = 0.001). There were no significant differences in secondary outcomes. Major periprocedural complication rate was 2.8%. Subset analysis on (i) cryoablation vs radiofrequency, (ii) Ensite vs CARTO navigational system and (iii) circular vs high density mapping catheter did not yield significant differences in primary or secondary outcomes. Conclusions The AF ablation complication and recurrence free rates in both paroxysmal and persistent AF at one year were comparable to high-volume centers. Long-term follow up is needed. In addition, first AF catheter ablation in a low-volume center is realistic with comparable efficacy and safety outcomes to high-volume centers using contemporary ablation technologies.

The blossom of online transaction has given rise to fraud, especially credit card fraud. Thus, Fraud Detection algorithm has become a critical issue, and several attempts have been made to detect transaction fraud using data mining methods. However, these methods suffer from lack of data or inadequate feature engineering. In this paper, we applied sufficient feature engineering and proposed a fraud detection algorithm based on Catboost. The experimental result indicates that our model outperforms other classical models, such as Logistic Regression, Support Vector Machine, and Random Forest. Moreover, we also illustrate the feature importance, which is valuable for feature selection and performance tuning.

Dysbiosis of gut microbiota has been linked to gestational diabetes mellitus (GDM), and grows as a resource for GDM biomarkers. However, the contributions of gut microbiota to GDM remain incompletely understood. Metabolites are key messengers in the interactions between gut microbiota and the host. Metabolomics is emerging as an essential tool in exploring the contributions of gut microbiota to diseases. In this study, we performed 1H-NMR based metabolomics on the feces of 62 pregnant women, including 31 women with GDM, and 31 women as the non-diabetes (NDM) control. Using Principle Component Analysis (PCA) and Orthogonal Projection to Latent Structures Discrimination Analysis (OPLS-DA), we observed clear cluster separation of the fecal metabolome between women with GDM and the NDM control. We further applied several feature selection methods to find five fecal metabolites contributing to the cluster separation of the fecal metabolome. These five metabolites, namely dibutyl decanedioate, N-acetylgalactosamine-4-sulphate, homocysteine, L-malic acid, and butanone, were significantly correlated with the clinical indices of GDM. Metabolite enrichment and pathway analysis on the five metabolites suggested that the fecal citrate cycle and sulfur metabolism were correlated with GDM. The results of this study demonstrated that disorders in the fecal metabolome are associated with GDM.

Most of smile recognition methods are based on constrained databases. Thus there are a lot of limitations when applying those algorithms into the real-world smile recognition. For the purpose of improving the accuracy in real-world smile recognition, we conducted our experiments on two databases (GENKI-4K database and our own built database). Depending on deep learning theory, we constructed a new deep model by stacking Contractive Auto-Encoder (CAE) on Contractive Denoising Auto-Encoder (CDAE) to extract useful features. Firstly, we pre-trained a CDAE to extract the feature of the first layer, then the extracted feature were used as input of the next basic model CAE, by pre-training the CAE model, we got more abstract feature, then the feature were used to classification. Experiments showed that our approach was useful for smile recognition. On the other hand, we also explored the influence of different number of training samples.

Introduction: Medtronic Ventricular Capture Management (VCM) allows effective long-term management of the pacing output and thus, maximises battery longevity. The aim of this study is to determine the accuracy of the VCM and results were comparable with in-office manual threshold results (MTR). Method: Patients with bradyarrhythmia indications who received a Medtronic device with VCM were recruited. Clinical demographics were collected and analysed. The VCM data was analysed and compared with in-office MTR. VCM was turned on only 3 months after implantation. Results: There were 30 patients (13 male, 17 female) of mean age 71 years old. 13% received single chamber pacemaker. The mean duration of initial implantation was 2.5 years. 14 patients (46.7%) had identical VCM threshold results and in-office MTR. 15 patients (50%) had+0.25 V difference between VCM threshold results and in-office MTR. Only 1 patient (3.3%) was shown to have 0.5 V difference between VCM threshold results (1.125 V @ 0.4 ms) and in-office MTR (0.625 V @ 0.4 ms). Conclusion: VCM is considered to be accurate and comparable to in-office manual threshold tests at least 3 months after device implantation. VCM reliability will allow proper management of the pacing output by adjusting accordingly to the threshold obtained automatically.