Search

In the past decade, the substantial achievements of therapeutic cancer vaccines have shed a new light on cancer immunotherapy. The major challenge for designing potent therapeutic cancer vaccines is to identify neoantigens capable of inducing sufficient immune responses, especially involving major histocompatibility complex (MHC)-II epitopes. However, most previous studies on T-cell epitopes were focused on either ligand binding or antigen presentation by MHC rather than the immunogenicity of T-cell epitopes. In order to better facilitate a therapeutic vaccine design, in this study, we propose a revolutionary new tool: a convolutional neural network model named FIONA (Flexible Immunogenicity Optimization Neural-network Architecture) trained on IEDB datasets. FIONA could accurately predict the epitopes presented by the given specific MHC-II subtypes, as well as their immunogenicity. By leveraging the human leukocyte antigen allele hierarchical encoding model together with peptide dense embedding fusion encoding, FIONA (with AUC = 0.94) outperforms several other tools in predicting epitopes presented by MHC-II subtypes in head-to-head comparison; moreover, FIONA has unprecedentedly incorporated the capacity to predict the immunogenicity of epitopes with MHC-II subtype specificity. Therefore, we developed a reliable pipeline to effectively predict CD4+ T-cell immune responses against cancer and infectious diseases. [ABSTRACT FROM AUTHOR]

• It is the first study to use deep learning to estimate SpO2 from face videos. • An appropriate combination of residual blocks and coordinate attention improves feature extraction effectively. • The color channel model (CCM) of calculating SpO2 is widely used and has good results. • Deep neural networks (DNN) can also estimate SpO2 directly from feature maps. • Combining the advantages of the CCM and the DNN, SpO2 estimation results are more accurate. Blood Oxygen (SpO 2), a key indicator of respiratory function, has received increasing attention during the COVID-19 pandemic. Clinical results show that patients with COVID-19 likely have distinct lower SpO 2 before the onset of significant symptoms. Aiming at the shortcomings of current methods for monitoring SpO 2 by face videos, this paper proposes a novel multi-model fusion method based on deep learning for SpO 2 estimation. The method includes the feature extraction network named Residuals and Coordinate Attention (RCA) and the multi-model fusion SpO 2 estimation module. The RCA network uses the residual block cascade and coordinate attention mechanism to focus on the correlation between feature channels and the location information of feature space. The multi-model fusion module includes the Color Channel Model (CCM) and the Network-Based Model(NBM). To fully use the color feature information in face videos, an image generator is constructed in the CCM to calculate SpO 2 by reconstructing the red and blue channel signals. Besides, to reduce the disturbance of other physiological signals, a novel two-part loss function is designed in the NBM. Given the complementarity of the features and models that CCM and NBM focus on, a Multi-Model Fusion Model(MMFM) is constructed. The experimental results on the PURE and VIPL-HR datasets show that three models meet the clinical requirement(the mean absolute error ⩽ 2%) and demonstrate that the multi-model fusion can fully exploit the SpO 2 features of face videos and improve the SpO 2 estimation performance. Our research achievements will facilitate applications in remote medicine and home health. [ABSTRACT FROM AUTHOR]

Abstract: It is of important to recognize the mechanical fault for extra-high voltage Circuit Breakers (CBs) in GIS, when the condition monitoring of CBs is realized. In this paper, a new efficiency fault recognition method is provided, by using improved Support Vector Machines (LibSVMs). The recognition methods, ANN and LibSVM are compared on their recognition accuracy, and the results show that the LibSVM is more efficient than ANN. The algorithm of LibSVM is improved by using Genetic Algorithm (GA), and the GA-LibSVM can obtain higher recognition accuracy than usual LibSVM for mechanical fault recognition of CB. [Copyright &y& Elsevier]

Abstract: This paper presents a new controller design technique for systems driven with impulse inputs. Necessary conditions for optimal impulse control are derived. A neural network structure to solve the resulting equations for optimal control is presented. Solution concepts are illustrated with example problems that exhibit increasing levels of difficulty. Two linear problems, one scalar and one vector and a benchmark nonlinear problem, the Van Der Pol oscillator, are used as case studies. Numerical results show the efficacy of the new solution process for impulse driven systems. Since the theoretical development and the design technique are free from restrictive assumptions, this technique is applicable to many problems in engineering and science. [Copyright &y& Elsevier]