Few-Shot Class-Incremental Learning (FSCIL) techniques are essential for developing Deep Learning (DL) models that can continuously learn new classes with limited samples while retaining existing knowledge. This capability is particularly crucial for DL-based retinal disease diagnosis system, where acquiring large annotated datasets is challenging, and disease phenotypes evolve over time. This paper introduces Re-FSCIL, a novel framework for Few-Shot Class-Incremental Retinal Disease Recognition (FSCIRDR). Re-FSCIL integrates the RETFound model with a fine-grained module, employing a forward-compatible training strategy to improve adaptability, supervised contrastive learning to enhance feature discrimination, and feature fusion for robust representation quality. We convert existing datasets into the FSCIL format and reproduce numerous representative FSCIL methods to create two new benchmarks, RFMiD38 and JSIEC39, specifically for FSCIRDR. Our experimental results demonstrate that Re-FSCIL achieves State-of-the-art (SOTA) performance, significantly surpassing existing FSCIL methods on these benchmarks

High-density surface electromyography (HD-sEMG) provides distinct advantages over traditional bipolar sEMG, including improved spatial resolution and enhanced localization of muscle activity. However, the collection of HD-sEMG signals requires the use of multiple electrodes over a small skin area, resulting in large data volumes. Managing such data necessitates substantial storage capacity, high bandwidth, and effective handling of information redundancy to ensure both usability and accuracy. Furthermore, challenges such as electrode contact variability and external electromagnetic interference often compromise signal quality, which calls for robust signal restoration techniques. This paper presents HD-sEMG-CORE, an open-source hybrid network algorithm designed for efficient compression and accurate restoration of HD-sEMG signals. The proposed approach combines generative adversarial networks (GANs) and variational autoencoders (VAEs) for signal compression, while a U-Net-based convolutional neural network (CNN) is employed to restore the features of corrupted or noisy signals in the latent space, completing the reconstruction process. Performance metrics, including mean squared error (MSE), mean absolute error (MAE), Pearson correlation coefficient (ρ), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM), demonstrate the effectiveness of HD-sEMG-CORE in both compression and restoration tasks. This methodology offers an efficient and precise solution for managing large HD-sEMG datasets, with potential applications in neurophysiology and neuroengineering.

Over the past two decades, the number of people with arm disabilities has dramatically increased. Many researchers have begun to concentrate on developing technologies to address this challenge. Exoskeleton is one of the most successful technologies out of these since it does not require medical staff to accompany the patients and can continually rehabilitate the patients until they regain the ability to move voluntarily. However, the majority of exoskeletons are anchored to the ground and individuals are unable to purchase one due to its high manufacturing costs. Furthermore, since this type of exoskeleton is anchored to the ground, patients must go to hospitals to get rehabilitation services. In this study, a portable assistive exoskeleton for human arm motions is designed to address this problem. This device can carry patients' arms to the desired position by adhering to a prescribed trajectory by medical specialists. The main parts of this exoskeleton are made of polylactic acid and produced using 3D printing technology. Therefore, the total manufacturing costs of the exoskeleton are not excessive and the weight of it is not high as well.

This study presents a novel method to tackle the difficulties in high-density surface electromyography (HD-sEMG) signal reconstruction: the electromyographic transformer generative adversarial network (E-TRGAN). Conventional techniques struggle with the intricate spatial-temporal dynamics of HD-sEMG and mostly concentrate on bipolar sEMG. E-TRGAN performs exceptionally well at reconstructing HD-sEMG signals while preserving their temporal and spatial integrity by utilizing the transformer model and the structure of generative adversarial networks (GANs).We present a thorough evaluation that includes comparisons with CNN-based methods and linear interpolation, showing that E-TRGAN performs better at recovering HD-sEMG signals in a variety of corruption conditions, such as reduced channels, random channel loss, and severe signal value loss. Metrics like MSE, MAE, RMSE, PSNR, and SSIM show that the model regularly performs better than competing approaches. Although E-TRGAN’s accuracy decreases in situations when there is simultaneous spatial and temporal signal loss, its overall efficacy represents a major breakthrough in HD-sEMG signal processing. This study highlights E-TRGAN’s promise as an adaptable approach for intricate HD-sEMG signal restoration in a range of applications.

Based on the principles of neuromechanics, human arm movements result from the dynamic interaction between the nervous, muscular, and skeletal systems. To develop an effective neural feedback controller for neuro-rehabilitation training, it is important to consider both the effects of muscles and skeletons. In this study, we designed a neuromechanics-based neural feedback controller for arm reaching movements. To achieve this, we first constructed a musculoskeletal arm model based on the actual biomechanical structure of the human arm. Subsequently, a hybrid neural feedback controller was developed that mimics the multifunctional areas of the human arm. The performance of this controller was then validated through numerical simulation experiments. The simulation results demonstrated a bell-shaped movement trajectory, consistent with the natural motion of human arm movements. Furthermore, the experiment testing the tracking ability of the controller revealed real-time errors within one millimeter, with the tensile force generated by the controller’s muscles being stable and maintained at a low value, thereby avoiding the issue of muscle strain that can occur due to excessive excitation during the neurorehabilitation process.

The neural control of human quiet stance remains controversial, with classic views suggesting a limited role of the brain and recent findings conversely indicating direct cortical control of muscles during upright posture. Conceptual neural feedback control models have been proposed and tested against experimental evidence. The most renowned model is the continuous impedance control model. However, when time delays are included in this model to simulate neural transmission, the continuous controller becomes unstable. Another model, the intermittent control model, assumes that the central nervous system (CNS) activates muscles intermittently, and not continuously, to counteract gravitational torque. In this study, a delayed reinforcement learning algorithm was developed to seek optimal control policy to balance a one-segment inverted pendulum model representing the human body. According to this approach, there was no a-priori strategy imposed on the controller but rather the optimal strategy emerged from the reward-based learning. The simulation results indicated that the optimal neural controller exhibits intermittent, and not continuous, characteristics, in agreement with the possibility that the CNS intermittently provides neural feedback torque to maintain an upright posture.