Search

Background: 5q spinal muscular atrophy (SMA) is a progressive autosomal recessive motor neuron disease. Objective: We aimed to assess the effects of nusinersen on motor function and electrophysiological parameters in adolescent and adult patients with 5q SMA. Methods: Patients with genetically confirmed 5q SMA were eligible for inclusion, and clinical data were collected at baseline (V1), 63 days (V4), 180 days (V5), and 300 days (V6). The efficacy of nusinersen was monitored by encompassing clinical assessments, including the Revised Upper Limb Module (RULM), Hammersmith Functional Motor Scale Expanded (HFMSE), 6-Minute Walk Test (6MWT), and percent-predicted Forced Vital Capacity in sitting position (FVC%) and Compound Muscle Action Potential (CMAP) amplitude. The patients were divided into "sitter" and "walker" subgroups according to motor function status. Results: 54 patients were screened, divided into "sitter" (N = 22) and "walker" (N = 32), with the mean age at baseline of 27.03 years (range 13–53 years). The HFMSE in the walker subgroup increased significantly from baseline to V4 (mean change +2.32-point, P = 0.004), V5 (+3.09, P = 0.004) and V6 (+4.21, P = 0.005). The patients in both the sitter and walker subgroup had no significant changes in mean RULM between V1 and the following time points. Significant increases in CMAP amplitudes were observed in both upper and lower limbs after treatment. Also, patients with RULM ≥ 36 points showed significant CMAP improvements. Our analysis predicted that patients with CMAP amplitudes of trapezius ≥ 1.76 mV were more likely to achieve significant motor function improvements. Conclusions: Nusinersen effectively improves motor function and electrophysiological data in adolescent and adult patients with SMA. This is the first report on the CMAP amplitude changes in the trapezius after treatment in patients with SMA. The CMAP values effectively compensate for the ceiling effect observed in the RULM, suggesting that CMAP could serve as an additional biomarker for evaluating treatment efficacy. [ABSTRACT FROM AUTHOR]

Although Graph Convolutional Networks (GCNs) have achieved excellent results in various graph-related tasks, their performance at low label rates is still unsatisfactory. Previous studies in Semi-Supervised Learning (SSL) for graph primarily focused on utilizing network predictions to generate pseudo-labels or instruct message propagation, often resulting in incorrect predictions owing to over-confidence. To address this issue, we propose a novel approach called Dual-Channel Consistency based Graph Convolutional Networks (DCC-GCN) for semi-supervised node classification. The key idea behind DCC-GCN is to leverage the extraction of embeddings from both node features and topological structures by employing GCN encoders in two separate channels. We consider samples with consistent predictions from these two channels as high-confidence samples, while those with differing predictions are labeled as low-confidence samples. DCC-GCN calibrate the feature embeddings of low-confidence samples by aggregating high-confidence samples from their respective neighborhoods. DCC-GCN can significantly improve the classification accuracy of low-confidence samples, thus improving the overall accuracy. Experiments on seven graph datasets demonstrate that DCC-GCN outperforms prior SSL methods, improving node classification accuracy by a considerable margin. [ABSTRACT FROM AUTHOR]

Face reenactment is a critical technology of digital face editing. Lately, the NeRFACE, a face reenactment method based on neural radiance fields, has been proposed, making the reconstruction accuracy of the training dataset much higher than the previous methods. However, face reenactment in realistic scenes often encounters poses and expressions that have not been seen before, which requires further improvement of the model's generalization capability. Based on the idea of ensemble learning, we present EnNeRFACE as using the adaptive ensemble neural radiance fields architecture, which is mainly composed of a set of subgenerators and a controller. We divide the short video of human portraits into non-intersecting sub-datasets based on time correlation, thus enabling the trained subgenerators to have differentiated modeling capabilities. In response to different expression vectors, the generator dynamically adjusts the weights assigned to each generator so that the capabilities of the subgenerators are adequately exploited. Extensive experiments show that EnNeRFACE has more stable and superior performance in generalization (i.e., identity preservation, manipulation of expression and pose) than the state-of-the-art methods, demonstrating the effectiveness of our proposed adaptive ensemble structure. [ABSTRACT FROM AUTHOR]

Talking face generation is widely used in education, entertainment, shopping, and other social practices. Existing methods focus on matching the speaker's mouth shape with the speech content. Still, there is a lack of research on automatically extracting potential head motion features from speech, resulting in a lack of naturalness. This paper proposes SATFace, a subject agnostic talking face generation method with natural head movement. To model the talking face's complicated and critical features (identity, background, mouth shape, head posture, etc.), we construct SATFace by taking encoder-decoder as the primary network architecture. Then, we design a long short-time feature learning network to better reference the global and local information in audio for generating reasonable head movement. Besides, a modular training process is proposed to improve explicit and implicit features' learning effects and efficiency. The experimental comparison results show that SATFace improves by at least about 9.8% in cumulative probability of blur detection and 8.2% in synchronization confidence compared with the mainstream methods. The mean opinion scores show that SATFace has advantages in terms of lip sync quality, head movement naturalness, and video realness. [ABSTRACT FROM AUTHOR]

Autoimmune nodopathy is a peripheral neuropathy characterized by acquired motor and sensory deficit with autoantibodies against the node of Ranvier or paranodal region in the peripheral nervous system. The clinical and pathological characteristics of the disease are distinct from that of chronic inflammatory demyelinating polyradiculoneuropathy (CIDP), and the standard treatment for CIDP is partially effective. Rituximab is a chimeric monoclonal antibody which binds and depletes B cells in peripheral blood. This prospective observational study included 19 patients with autoimmune nodopathy. Participants received intravenous rituximab treatment 100 mg the first day and 500 mg the next day and given every 6 months. The Inflammatory Neuropathy Cause and Treatment (INCAT) disability score, Inflammatory Rasch-Built Overall Disability Scale (I-RODS), Medical Research Council (MRC) sum score, and Neuropathy Impairment Score (NIS) were collected at entry and before the rituximab infusion every 6 months. At the last visit, 94.7% (18/19) of the patients showed clinical improvement on either the INCAT, I-RODS, MRC, or NIS scale. After the first infusion, 9 patients (47.7%) showed improvement on the INCAT score, and 11 patients (57.9%) on cI-RODS. In patients who received more than one rituximab infusion, the improvement of INCAT score and cI-RODS at the last assessment was higher than that after the first infusion. We also observed tapered or withdrawn concomitant oral medications in these patients. [ABSTRACT FROM AUTHOR]

Inspired by recent success of graph contrastive learning methods, we propose a self-supervised learning framework for Graph Neural Networks (GNNs) named Adaptive Multi-layer Contrastive Graph Neural Networks (AMC-GNN). Specifically, AMC-GNN generates different graph views through data augmentation and compares the output embeddings at different layers of graph neural network encoders to obtain feature representations for downstream tasks. Meanwhile, AMC-GNN learns the importance weights of embeddings at different layers adaptively through the attention mechanism, and an auxiliary encoder is adopted to train graph contrastive encoders better. The accuracy is improved by maximizing the representation's consistency of positive pairs in the intermediate layers and the final embedding space. Experiments on node classification and link prediction demonstrate that the AMC-GNN framework outperforms state-of-the-art contrastive learning methods and even sometimes outperforms supervised methods. [ABSTRACT FROM AUTHOR]

The Heilongjiang Jianbiannongchang area is located at the confluence of the Great and Lesser Xing'an Ranges. This area has a complex magmatic and tectonic evolutionary history that has resulted in a complex and diverse geological background for mineralization. In this study, isometric logarithmic ratio (ILR) transformations of Au, Cu, Pb, Zn, and Sb contents were performed in the 1:50,000 soil geochemical data of the Jianbiannongchang area. Robust principal component analysis (RPCA) was conducted based on ILR transformation. The local singularity and spectrum-area (S-A) methods were used to extract information on mineralogic anomalies. The results showed that: (1) the transformed data eliminated the influence of the original data closure effect, and the PC1 and PC2 information obtained by applying RPCA reflected ore-producing element anomalies dominated by Au and Cu. (2) The local singularity method can enhance the information of the local strong and weak slow anomalies. After performing local singularity analysis on PC1 and PC2, the obtained local anomalies reflected the local singularity spatial anomaly patterns related to Cu and Au mineralization in this area, which is an effective method for trapping ore-producing anomalies. (3) Furthermore, the composite anomaly decomposition of PC1 and PC2 was performed using the S-A method, and the screened anomalous and background fields reflect the ore-producing anomalies related to Cu and Au mineralization. This information is in agreement with known Cu and Au mineralization. (4) The geochemical anomalies with mineralization potential were obtained outside the known mineralization sites by integrating the information of ore-producing anomalies extracted by the local singularity and S-A methods, providing the theoretical basis and exploration direction for future exploration in the study area. [ABSTRACT FROM AUTHOR]

The design requirements of the Fourier transform optical system required for synthetic aperture radar(SAR) are analyzed and the specific optical parameters are determined. The design of the Fourier lens adopts a fully symmetrical structure to avoid simultaneous optimization of the forward and reverse directions of the optical path and can complete the correction of the conjugate positional aberrations of the two pairs of object images. A Fourier lens composed of four spherical mirrors was designed and reached the diffraction limit. To verify that the optical lens can realize the Fourier transform, a 4f optical system was built in the laboratory using the lens. First, the 4f system built by the optical element was demonstrated to be able to process the signal by demodulating a one-dimensional linear frequency modulation(LFM) signal. Then, the echo signal of the SAR point target is demodulated, Fourier transform is realized by the lens, and the imaging of SAR single point target signal is completed, which plays a role in the practical application of the fully symmetrical Fourier optical lens. [ABSTRACT FROM AUTHOR]

Although the viscosity behavior of bacteria and extracellular polymeric substances (EPS) in flocculent activated sludge (FAS) and aerobic granular sludge (AGS) has been investigated, no studies have explored the role of viscosity in microbial attachment in pure culture. This study investigated the viscosity behavior of bacteria and EPS. The results showed that bacteria and their EPS exhibited non-Newtonian fluid and shear-thinning behavior. The viscosity of bacteria and EPS was 1.55–3.80 cP and 1.10–2.40 cP, respectively, while the attachment of bacteria (optical density at 600 nm) was 0.1426–3.1015. Bacteria with high attachment secreted EPS with a higher viscosity (2.40 cP), whereas those with weak attachment expressed EPS with a lower viscosity (1.10 cP). Viscosity and microbial attachment or extracellular polysaccharide (PS) content were significantly positively correlated. PS content was the source of bacterial viscosity, and β-polysaccharide played a more important role in viscosity and microbial attachment than α-polysaccharide. Thus, viscosity plays a critical role in microbial attachment, and high viscosity and PS content result in high microbial attachment, which is beneficial to the granulation process of AGS. [ABSTRACT FROM AUTHOR]

Large labeled data bring significant performance improvement, but acquiring labeled medical data is particularly challenging due to the laborious, time-consuming, and medically qualified annotation. Semi-supervised learning has been employed to leverage unlabeled data. However, the quality and quantity of annotated data have a great influence on the performance of the semi-supervised model. Selecting informative samples through active learning is crucial and could improve model performance. Therefore, we propose a unified semi-supervised active learning architecture (RL-based SSAL) that alternately trains a semi-supervised network and performs active sample selection. Semi-supervised model is first well trained for sample selection, and selected label-required samples are annotated and added to the previously labeled dataset for subsequent semi-supervised model training. To learn to select the most informative samples, we adopt a policy learning-based approach that treats sample selection as a decision-making process. A novel reward function based on the product of predictive confidence and uncertainty is designed, aiming to select samples with high confidence and uncertainty. Comparisons with a semi-supervised baseline on collected lumbar disc herniation dataset demonstrate the effectiveness of the proposed RL-based SSAL, achieving over 3% promotion across different amounts of labeled data. Comparisons with other active learning methods and ablation studies reveal the superiority of proposed policy learning based on active sample selection and reward function. Our model trained with only 200 labeled data achieves an accuracy of 89.32% which is comparable to the performance achieved with the entire labeled dataset, demonstrating its significant advantage., Competing Interests: Declarations. Ethics Approval: This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Ethics Review Committees of the Fifth Affiliated Hospital of Sun Yat-sen University (No. 2021 K53-1). Consent to Participate: The study was done after agreement from the hospital ethics committee and with the informed consent of the individual participants. Consent for Publication: The work is original and not previously published. All authors agreed to submit this manuscript to this journal. The authors affirm that human research participants provided informed consent for publication. Competing Interests: The authors declare no competing interests., (© 2024. The Author(s) under exclusive licence to Society for Imaging Informatics in Medicine.)