Advanced intelligent systems (Weinheim an der Bergstrasse, Germany)最新文献

Accurate single-cell analysis is critical for diagnostics, immunomonitoring, and cell therapy, but coincident events, where multiple cells overlap in a sensing zone, can severely compromise signal fidelity. A hybrid framework combining a fully convolutional neural network (FCN) with compressive sensing (CS) to disentangle such overlapping events in 1D sensor data is presented. The FCN, trained on bead-derived datasets, accurately estimates coincident event counts and generalizes to immunomagnetically labeled CD4⁺ and CD14⁺ cells in whole blood without retraining. Using this count, the CS module reconstructs individual signal components with high fidelity, enabling precise recovery of single-cell features, including velocity, amplitude, and hydrodynamic diameter. Benchmarking against conventional state-machine algorithms shows superior performance, recovering up to 21% more events and improving classification accuracy beyond 97%. Explainability via class activation maps and parameterized Gaussian template fitting ensures transparency and clinical interpretability. Demonstrated with magnetic flow cytometry (MFC), the framework is compatible with other waveform-generating modalities, including impedance cytometry, nanopore, and resistive pulse sensing. This work lays the foundation for next-generation nonoptical single-cell sensing platforms that are automated, generalizable, and capable of resolving overlapping events, broadening the utility of cytometry in translational medicine and precision diagnostics, e.g., cell-interaction studies.

As the most common pediatric malignancy, B-cell acute lymphoblastic leukemia (B-ALL) has multiple distinct subtypes characterized by recurrent and sporadic somatic and germline genetic alterations. Identifying B-ALL subtypes can facilitate risk stratification and enable tailored therapeutic design. Existing methods for B-ALL subtyping primarily depend on immunophenotyping, cytogenetic tests, and genomic profiling, which can be costly, complicated, and laborious. To overcome these challenges, RanBALL (an ensemble random projection-based model for identifying B-ALL subtypes) is presented, an accurate and cost-effective model for B-ALL subtype identification. By leveraging random projection (RP) and ensemble learning, RanBALL can preserve patient-to-patient distances after dimension reduction and yield robustly accurate classification performance for B-ALL subtyping. Benchmarking results based on >1700 B-ALL patients demonstrate that RanBALL achieves remarkable performance (accuracy: 0.93, F1-score: 0.93, and Matthews correlation coefficient: 0.93), significantly outperforming state-of-the-art methods like ALLSorts in terms of all performance metrics. In addition, RanBALL performs better than t-SNE in terms of visualizing B-ALL subtype information. We believe RanBALL will facilitate the discovery of B-ALL subtype-specific marker genes and therapeutic targets to have consequential positive impacts on downstream risk stratification and tailored treatment design is believed. To extend its applicability and impacts, a Python-based RanBALL package is available at https://github.com/wan-mlab/RanBALL.

Soft robotics requires compliant actuators for safe human interaction. While McKibben artificial muscles are popular for their high force output, their reliance on bulky, noisy pumps limits their use in wearable devices. Electrohydrodynamic (EHD) pumps offer a compact and silent alternative, but existing designs struggle with dielectric discharge and fabrication issues, which compromise reliability and power density. This study introduces a novel EHD pump featuring 0.1 mm copper wire electrodes in a diagonal arrangement within a laser-cut acrylic frame. This design improves dielectric resilience, minimizes deformation, and allows for compact integration. A new simplified fabrication process results in sample variation under 5%. The pump demonstrates remarkable performance, achieving 107 kPa pressure and an 88 mL min⁻¹ flowrate, doubling the power density of the previous model while retaining 88% of its flowrate after 50 discharge events. An automated self-recovery mechanism is also implemented, enabling the pump to instantly restore function after a discharge. When paired with a McKibben muscle, the system achieves a 2 s contraction time, a tenfold improvement over the prior EHD-driven system. This work presents a significant advancement in fast, resilient, and scalable actuation, paving the way for next-generation wearable robotics and assistive technologies.

Macrophages play a central role in modulating different biological and physiological events. The behaviors and functions of macrophages may be regulated by a host of factors, including their viability, proliferation rate, and population density. Specifically, the population density of macrophages has been increasingly reported to be correlated with their activities. It is, however, still unclear if changes in macrophage population density will alter the biophysical attributes of these cells, notably their morphology. Herein, label-free phase-contrast microscopy is coupled with machine learning to interrogate the relationship between the population density and morphological features of macrophages. Through a systematic approach, variations in the morphological phenotypes of macrophages, which are dependent on their population density, are revealed. In parallel, through unsupervised clustering, the presence of single-cell morphological heterogeneity within each macrophage population and subpopulation is elucidated. Next, discriminative morphological attributes which can be leveraged to distinguish between macrophages from different groups are identified through feature scoring. Finally, high-performing explainable supervised machine learning algorithms that can be employed to predict the population density of macrophages based on their size and shape features are identified. This work is anticipated to offer a deeper understanding of the association between macrophage population density and morphologyas well as the potential use of morphological attributes as predictive metrics for analyzing cell populations.

Electrical stimulation and microneedle-mediated drug delivery emerge as promising therapies in gastrointestinal (GI) motility disorders and inflammatory conditions. However, on-demand intervention therapy in enclosed narrow GI remains a challenge. Herein, a magnetic-driven soft membrane robot is presented that synergistically combines microneedle-mediated electrical stimulation and drug delivery. The membrane robot's bipolar magnetization enables switching between two surfaces by external magnetic fields, where N-pole drives treatment surface with microneedle to penetrate GI wall and S-pole initiates smooth surface for low resistance locomotion. The membrane robot utilizes magnetically coupled resonant wireless transmission to enable regulated electrical stimulation with 86.7% efficiency at 6 cm distance, while providing tunable voltage (0–20 V) and programmable pulse waveforms (0.4–50 ms width) for adaptive bioelectrical modulation. The drug-loaded microneedle array serves dual roles as both a penetrating electrode and a therapeutic interface, delivering electrical stimulation while simultaneously releasing encapsulated agents upon tissue penetration. In vitro experiments of the multimode motion and multifunctional treatment are validated in a fresh pig gut. This integrated membrane magnetic robot offers great potential in GI diagnostics, personalized neuromodulation, and on-demand drug release applications.

Dysarthric speech recognition faces significant challenges of acoustic variability and data scarcity, and this study proposes a robust system by integrating generative adversarial network enhancement and large language model correction to address these issues effectively. The system employs three key components, including a multimodal recognition core that combines whisper-medium encoder with LoRA-fine-tuned Llama-3.1-8B for end-to-end acoustic-to-semantic mapping, an improved CycleGAN module that generates synthetic dysarthric speech through Inception-ResNet fusion blocks, and an intelligent error correction mechanism using N-best hypothesis reranking with semantic constraints. Experiments on the UA-Speech dataset show that the complete system achieves a 20.61% word error rate representing a 73.9% relative improvement over traditional end-to-end transformer automatic speech recognition. Under very low intelligibility conditions it maintains a 48.69% word error rate demonstrating robust recognition for severe pathological speech. Ablation studies validate each module's effectiveness, providing significant advances for dysarthric patient communication technologies.

This article proposes a neural network-based parameter extraction methodology for the Berkeley Short-Channel IGFET Model–Common Multi-Gate (BSIM–CMG) model applied to gate-all-around field effect transistors (GAAFETs), capturing both current–voltage and capacitance–voltage characteristics to support compact model library development. Conventional BSIM parameter extraction is often complex and inefficient, requiring manual intervention and significant time to cover a wide range of device dimensions and temperatures. To address these limitations, a novel binning adaptive sampling strategy is integrated into the neural network-based extraction framework to efficiently generate training data across broad device dimension ranges. In addition, the transformer-based deep neural networks are designed to output only binnable parameters, ensuring compatibility with compact model library requirements. The trained networks are tested using 3 nm node GAAFET Technology Computer Aided Design (TCAD) data under various conditions, achieving mean absolute percentage errors below 5% for both drain current and gate capacitance. Consequently, the extracted parameters are integrated with corner model parameters through binning equations. This approach results in binning models that are readily deployable in compact model libraries while significantly reducing parameter extraction time and enabling automation across a wide range of GAAFET dimensions.

With the advancement of artificial intelligence (AI), there is an increasing demand for high-speed, energy-efficient hardware capable of running complex machine learning algorithms. Traditional hardware is constrained by the Von Neumann bottleneck, resulting in high power consumption and slower speeds. Inspired by the human brain, bio-mimicking the dynamic synaptic plasticity of the biological synapse using synaptic transistors is crucial to building the next generation of high-performance computing hardware-based neural networks. This study investigates neuromorphic behavior in 180 nm bulk complementary metal oxide semiconductor (CMOS) devices at 4 K, emphasizing memory properties and synapse-like characteristics. These findings position bulk CMOS as a scalable, energy-efficient, cryo-compatible platform for neuromorphic and quantum computing use. Gated-pulse measurements are used to study potentiation–depression behavior by quantifying conductance changes as functions of pulse amplitude and width. These results closely resemble biological synaptic plasticity, laying the groundwork for integrating cryo-CMOS technology into neuromorphic computing. The work reported here aims to work toward the development of hybrid computational systems by bridging the gap between conventional CMOS devices and emerging cryogenic technology, offering new avenues for scalable, energy-efficient, and high-performance cryogenic neuromorphic technologies.

Animal-Robot Interaction

The cover illustrates a gregarious locust interacting with a biomimetic agent inoculated with Beauveria bassiana on an robotic experimental platform, highlighting the dynamics of social immunity and pathogen information spread within the swarm, as explored through innovative biohybrid method of this study. More details can be found in article 2400763 by Donato Romano and Cesare Stefanini.

Siamese Network

This paper proposed a novel AI framework that enhances guidewire tip tracking in image-guided therapy for vascular diseases. Combining a Siamese network with attention mechanisms ensures robust tracking despite visual ambiguities and tissue deformations. Validated on clinical angiography sequences and robotic platforms, it improves diagnostic and therapeutic precision in endovascular interventions. More details can be found in the Research Article by Peng Qi and co-workers (DOI: 10.1002/aisy.202500425).