Search

In recent years, the sample mining strategy has been integrated into the loss function of face recognition, significantly improving the performance of face recognition. But most of the work focuses on how to mine difficult samples during the training phase, without considering the potential unrecognized sample images in the difficult samples, resulting in poor recognition performance of the model for low-quality facial images. To solve this problem, this paper proposes a hybrid adaptive loss function MixFace that combines sample difficulty adaptation and image quality adaptation. The loss function combines the CurricularFace based on curriculum learning with the image adaptive loss function AdaFace. The feature norm is incorporated into the loss function as an image quality indicator. On the premise of focusing on image quality, this paper focuses on simple samples in the early training stage and difficult samples in the later training stage, reducing the network model’s attention to some low-quality unrecognized samples in difficult samples. Trained on CASIA-WebFace and MS1MV2 datasets, MixFace shows varying degrees of performance improvement compared with CurricularFace and AdaFace on high-quality test sets LFW, CFP_FP, AgeDB, CALFW, and CPLFW. At the same time, MixFace shows better recognition performance than CurricularFace and AdaFace on medium quality test sets IJB-B, IJB-C and low-quality test set TinyFace. Experimental results show that MixFace can effectively reduce the interference of unrecognized images, thereby improving the performance of low-quality face recognition. At the same time, benefiting from the curriculum learning method in MixFace, it can still maintain good performance for high-quality face recognition.

Wet age-related macular degeneration(wARMD)emerges as a primary contributor to irreversible vision impairment in the aging demographic. In clinical practice, anti-vascular endothelial growth factor(VEGF)therapies exhibit pronounced success in managing wARMD. However, in the actual clinical application, there are significant individual differences in the prognosis of anti-VEGF drug therapy, and some patients show poor response to the treatment, which may be related to the morphological differences of retinal layers in macular area, genetics, systemic conditions and other factors. It will help develop a more rational and individualized treatment plan to judge the prognosis of patients according to their different clinical manifestations in advance, so as to reduce overtreatment and the risk of retinal damage. In recent years, most studies on treatment response mainly focus on fundus morphology, genetics and so on. In this study, the relevant factors affecting adverse response to wARMD were reviewed, aiming to provide with more accurate treatment and prognostic monitoring programs for clinicians.

PurposeWe evaluated whether combining a humanized antidisialoganglioside monoclonal antibody (hu14.18K322A) throughout therapy improves early response and outcomes in children with newly diagnosed high-risk neuroblastoma.Patients and methodsWe conducted a prospective, single-arm, three-stage, phase II clinical trial. Six cycles of induction chemotherapy were coadministered with hu14.18K322A, granulocyte-macrophage colony-stimulating factor (GM-CSF), and low-dose interleukin-2 (IL-2). The consolidation regimen included busulfan and melphalan. When available, an additional cycle of parent-derived natural killer cells with hu14.18K322A was administered during consolidation (n = 31). Radiation therapy was administered at the end of consolidation. Postconsolidation treatment included hu14.18K322A, GM-CSF, IL-2, and isotretinoin. Early response was assessed after the first two cycles of induction therapy. End-of-induction response, event-free survival (EFS), and overall survival (OS) were evaluated.ResultsSixty-four patients received hu14.18K322A with induction chemotherapy. This regimen was well tolerated, with continuous infusion narcotics. Partial responses (PRs) or better after the first two chemoimmunotherapy cycles occurred in 42 of 63 evaluable patients (66.7%; 95% CI, 55.0 to 78.3). Primary tumor volume decreased by a median of 75% (range, 100% [complete disappearance]-5% growth). Median peak hu14.18K322A serum levels in cycle one correlated with early response to therapy (P = .0154, one-sided t-test). Sixty of 62 patients (97%) had an end-of-induction partial response or better. No patients experienced progressive disease during induction. The 3-year EFS was 73.7% (95% CI, 60.0 to 83.4), and the OS was 86.0% (95% CI, 73.8 to 92.8), respectively.ConclusionAdding hu14.18K322A to induction chemotherapy improved early objective responses, significantly reduced tumor volumes in most patients, improved end-of-induction response rates, and yielded an encouraging 3-year EFS. These results, if validated in a larger study, may be practice changing.

Aiming at the problems that the facial features extracted from current face recognition algorithms are not distinguishable and the discrimination of difficult and easy samples is not enough, a face recognition algorithm combining attention mechanism and curriculum learning is proposed, which is called efficient cooperative attention and curriculum face (ECACFace). The algorithm proposes an efficient spatial channel attention module (ESCA) and integrates it into the basic module of the feature extraction network. The efficient channel attention module (ECA) is used to obtain the channel attention, and the spatial attention module is added after the ECA. On the basis of paying attention to the image channel information, the spatial attention is further obtained, and the face feature vector with richer information is obtained for face classification. At the same time, the loss function based on curriculum learning is introduced to distinguish the difficult and easy samples in the training process. The simple samples are trained in the early stage and the difficult samples are trained in the later stage to realize the discriminative sample learning. Training ECACFace based on lightweight network and shallow network on CASIA-WebFace dataset and it has an accuracy improvement of more than 1.5 percentage points compared with the original network. ECACFace based on deep network is trained on MS1MV2 dataset which has millions of data, and the accuracy tested on CPLFW dataset is increased by 1.14 percentage points compared with ArcFace. Experimental results show that the cooperation of ESCA module and the loss function based on curriculum learning can further improve the perfor-mance of face recognition.

Considering the operation efficiency of the metro system, the boarding queuing problem is a complex and intractable dilemma caused by the security and ticket checking process in urban metro stations, especially for the mass-transit metro system in China. In this study, a novel scheme of fare differentials based on demand control and congestion management for reducing congestion in the metro boarding process was proposed. In addition, we extended Vickrey's point-queue model into a boarding congestion model incorporating the bidesired departure time considering the security and ticket checking process in stations as a continuum bottleneck. The train-run and departure time choosing behaviors of passengers under different schedule gaps of the adjacent train runs are explored when the boarding system achieves user equilibrium in the proposed model. Then, we examined the demand regulatory mechanisms of fare incentives (fare differentials) in reducing the queuing boarding time in metro stations when the optimization of the schedule gaps is ineffective in the pattern of mass-scale travel demand. The analytical solutions of these two optimal methods (schedule gaps and fare differentials) for boarding congestion management are presented. After comparing the two congestion-reduced methods, the fare incentive rule has a better regulation effect on the smoothing of travel demand. The results of the sensitivity analysis using numerical simulations reveal the regulatory mechanism of fare differentials in reducing the queuing time and increasing the incremental revenue. (1) The demand threshold is only related to the boarding capacity and schedule gaps (i.e., the greater the boarding capacity and schedule gaps, the greater the passenger capacity of the metro station). (2) The effect of fare incentives in reducing the boarding congestion is better if the lower fare is implemented in the later train runs. (3) A lower fare differential between two adjacent metro runs can be used to regulate the proportion of staggered passengers in the queuing line to reduce the crowd gathering in the metro station hall when travel demand is high, meanwhile, a higher fare differential between two adjacent metro shuttles can increase incremental revenue effectively. (4) The measure of fare differentials causes worse results in both the reduction of the queuing time and the increase of the incremental revenue when the metro travel demand is lower than the demand threshold. This conclusion is consistent with the pattern of reality and experience. Therefore, the definition and judgment conditions of demand thresholds, introduced in this study, can provide theoretical guidance when implementing fare incentive policies in aviation and metro networks., Author(s): Haitao Pan [1]; Shumin Feng (corresponding author) [1] 1. Introduction With the rapid increase in travel demands in urban areas, high passenger flow becomes a common phenomenon in the [...]

The regulatory standards of the United States Food and Drug Administration (FDA) require substantial evidence of effectiveness from adequate and well-controlled trials, for drugs developed in both adults and children. However, when scientifically justified, relying on extrapolation may be acceptable. Historically, the FDA's extrapolation approach was based on draft guidance published in 2014, which introduced the categories of full, partial, and no extrapolation. The European Medicines Agency (EMA) took a different view on pediatric extrapolation. To better understand the use of extrapolation to support pediatric drug development and approval, we reviewed the pediatric labeling changes published by the FDA, focusing on the labeling updates between 1/1/2015 and 7/31/2021, the period where the extrapolation approach is in transition to harmonize with the EMA. Within this time window, among the 265 drugs and biological products with pediatric labeling changes, 169 (63.8%) were identified where extrapolation was used. This includes 64 (24.2%) labeling changes, where full extrapolation was used, and 105 (39.6%) labeling changes, where partial extrapolation was used. The major disease areas that extrapolation was used include neuroscience (40/53, 75.5%) and infectious disease (20/28, 71.4%). The extrapolation approach was identified in terms of source population beyond the use of adult as well as extrapolation from clinical trials conducted in the same drug class. The use of extrapolation increased the rates of new and expanded pediatric indication in the period. This review gives the most recent landscape of pediatric labeling changes using extrapolation. With the released ICH (International Council for Harmonization) E11A guidance in April 2022, the paper also provides insights for future pediatric drug development programs. [ABSTRACT FROM AUTHOR]