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

Multisensory integration (MSI) is a core brain function underlying perception, learning, and behavior. Understanding the computational mechanisms of MSI is key to advancing AI and brain-inspired systems. While earlier models relied on probabilistic frameworks, recurrent neural networks (RNNs) offer advantages in capturing temporal dynamics and neural computations. This review presents a critical examination of computational models of MSI, focusing on the evolution from probabilistic integration to modern RNN-based methods. Biological evidence for temporal coordination in multisensory areas is analyzed and explored how different RNN architectures (e.g., vanilla, long short-term memory, and gated recurrent unit) simulate these dynamics. Comparative analyses show RNNs’ superiority in robustness and learning efficiency, with up to 46.9% improvement in classification tasks involving sensory fusion. We introduce a taxonomy of MSI tasks and a novel evaluation framework for model benchmarking. Real-world case studies—from speech recognition to prosthetic control—highlight practical applications. Challenges in interpretability, data efficiency, and generalization are also discussed. The review provides actionable insights for future research in both computational neuroscience and artificial intelligence. By bridging neurobiological principles and machine learning, RNN-based models pave the way for intelligent systems capable of flexible, context-aware multisensory processing.

The treatment plan for brain tumors varies depending on the type and stage of the tumor. Early diagnosis plays a vital role in determining appropriate treatment. In addition to clinical routines, artificial intelligence-based systems that produce automated, quantitative, and objective results can assist clinicians and scientists in making early diagnoses. For this motivation, this study proposes a deep learning-based system that classifies brain tumors obtained by magnetic resonance imaging. In the proposed approach, several wavelet transform approaches are applied to the raw dataset images. Thus, in addition to automated feature extraction in deep learning, it aimed to detect more detailed features. Therefore, four types of datasets have been obtained. Then, using the transfer learning approach, some popular convolutional neural network and vision transformer models are trained separately with the four-type datasets, and the test results are compared. The networks that produced the highest results are used to make the final decision with the ensemble technique. In the first analysis, the best performance was obtained using original data with an 83.50% accuracy value, and the second highest performance is obtained 81.72% accuracy value using the Daubhecies wavelet before deep learning. The third and fourth high performances are 81.47% and 81.22% accuracy, respectively, using original data. In the ensemble analysis, the highest result is achieved at 85.03% accuracy value using the bagging-ensemble approach of the networks, namely MobileNet-v3, vision transformer, ResNeXt, and DenseNet-201. This study demonstrates that using a hybrid wavelet transform and deep learning approach improves classification performance. This may inspire the use of the same method to solve different classification problems.

Kidney stones can cause severe pain and complications like chronic kidney disease. Although retrograde intrarenal surgery with laser lithotripsy is effective, current retrieval methods are inefficient, typically capturing only 1–3 fragments per ureteroscope pass and requiring many passes for full clearance. A novel spinner device that enables ultra-efficient fragment removal through spinning-induced localized suction is introduced. It generates spiral and circulating flows to capture fragments from over 20 mm away, eliminating the need to chase them. Optimized via computational fluid dynamics and validated in vitro and ex vivo, the spinner retrieves ≈60 small (0.5–2 mm) or ≈15 larger (2–3 mm) fragments per pass. It demonstrates nearly 100% capture of 60 fragments in bench tests and removes 45 fragments in 4 s in a porcine kidney model. This technology markedly improves procedural efficiency by reducing operative time, increasing stone-free rates, and minimizing the number of ureteroscope passes.

Continuum manipulators can operate in complex environments where traditional rigid manipulators fail. However, the modeling of inverse kinematics remains challenging because of its inherent nonlinearities and various external conditions. This work proposes an online learning control framework with a data cache pool utilizing a constant-curvature model inspired neural network (CCMINN) model to obtain the inverse kinematics model of tendon-actuated continuum manipulators. The CCMINN model is a kind of geometric relationships-inspired neural network, which is inspired by the geometric relationships within the constant-curvature model. This model improves the ability of traditional fully connected neural network models on high convergence speed and precision through its constant-curvature inspiration layers. These layers embed geometry insights into the neural network structure rather than loss functions like physics-informed neural networks. The online learning framework enables CCMINN to maintain high control accuracy in a variety of external load scenarios. Experiments show average tracking errors of 1.4 mm, 1.38 mm, and 1.48 mm (0.7%, 0.64%, and 0.74% of the continuum manipulator length) in the free space, under constant and variable loading conditions, respectively. The results show that combining the fast-converging CCMINN with an online learning control framework enables high-precision and robust positioning control of continuum manipulators under various external payloads.

Soft and continuum robots commonly rely on fluid, tendon, or rod-based power transmissions, to control robotic form and actuation. Architected geometry has enhanced robot control through tailored physical and mechanical properties based on topology. For example, twist-actuated metamaterials, such as handed shearing auxetics (HSAs), have expanded the soft robot design space, offering varied shape changes and direct integration with simple motors. Despite these advancements, current options for torque transmission limit the successful integration of HSAs in tubular robots, especially for constructions requiring maximized interior space for passing devices or additional concentric tubes. An architected structure based on patterned straight-line mechanisms is proposed that enables simultaneous bending, extending, and torsionally rigid (BETR) transmission. Pairing these new torque-transmitting materials with twist-driven materials HSAs creates new modalities for the varied actuation of tubular robots. Parameter trade offs in BETRs are analyzed, and a user operated robot is built that demonstrates feasibility for navigation, positioning, and anchoring in scaled 3D-printed anatomies.

Periodontal disease (PD) is a chronic condition associated with systemic risks like cardiovascular disease and diabetes. Traditional diagnostics detect advanced PD but often miss early-stage cases, where timely intervention is critical. Oral polymorphonuclear neutrophils (oPMNs) are emerging as key biomarkers for periodontal health. This study presents DePerio, an AI-driven deep neural network (DNN) method that isolates and quantifies oPMNs from saliva using their natural hydrophilic adhesion on treated surfaces. Trained on thousands of annotated bright-field images, DePerio accurately detects and counts oPMNs within milliseconds. Validation against standard techniques confirms its precision in measuring oral inflammatory load (OIL). Clinical testing on 51 samples from healthy and periodontitis patients demonstrates DePerio's capability to distinguish five OIL levels, assisting in PD severity assessment. This low-complexity, AI-powered tool offers a rapid, reliable approach for early PD detection and management in dental practices.

Mechanical resilience is crucial for both animals and machines. Repairing or replacing damaged components of machines is often costly and time-consuming. Many walking insects, especially species that autotomize legs as a predator-avoidance strategy, exhibit remarkable adaptive control of their leg movement dynamics to compensate for leg loss. The embodied adaptation of leg control in insects can be informative for robotics to develop control strategies for damage compensation. From this point, the study utilizes the stick insect Medauroidea extradentata as a model organism to investigate the effects of leg amputation on the compensatory control of walking behavior. A decentralized adaptive resilient neural control system is proposed, leveraging self-embodied resilience strategies, for legged robots. Unlike model-based or machine learning-based approaches, relying on accurate mathematical models or extensive training data, the neural control system achieves self-organized gait patterns and adaptive leg movements through minimal sensory feedback, coupled with neural dynamics, synaptic plasticity, and robot-environment interactions. This embodied neural control approach is validated and demonstrated on simulated and real insect robots, resulting in robust locomotion and rapid adaptation (within seconds) to various leg loss cases. The combined findings reveal the potential for insect-inspired embodied emergent resilience in complex robotic systems toward resilient robotics.

Correct and rapid classification of brain tumor types is crucial for the patient's treatment plans. This study aims to create a deep learning-based mobile application that leverages on-device AI capabilities to classify brain tumors. For this reason, first, a series of preprocessing steps are applied to MR images. Then, convolutional neural network , ViT, and MobileViT models are trained for this task. Also, pretrained VGG16, ResNet152V2, InceptionV3, InceptionResNetV2, and MobileNetV2 models are retrained for the brain tumor classification task with the transfer learning method. Using the publicly available “Brain Tumor MRI Dataset,” the model performances are evaluated, and test results are compared. MobileViT shows the best performance in terms of balance between inference time and success rate. Thus, the TensorFlow model of MobileViT is converted to the TensorFlow Lite model and integrated into the mobile application. The mobile application is developed using the Flutter framework. The application has been evaluated on two different devices, and 298.98 and 317.50 ms average inference times have been observed. The proposed system shows that rapid and effective brain tumor classification can be performed by integrating deep learning into the mobile application. This system can assist experts in the decision-making process.

This paper proposes a novel hybrid framework to accurately identify human peripheral blood cells. Our approach includes Big Transfer (BiT) models, combining the extracted features with classifiers: the traditional Multilayer Perceptron (MLP), the Efficient Kolmogorov-Arnold Network (EfficientKAN) and our hybrid method (HybridMLPEfficientKAN). Peripheral Blood Cell (PBC) dataset of 17092 images covering eight cell types is preferred. BiT models provide high-dimensional features for classifications pipelines. Results show that combining MLP and EfficientKAN provides strong classification accuracy while reducing training overhead often seen in standalone EfficientKAN. Training durations in HybridMLPEfficientKAN remain close to MLP training, in the range of 100-250 seconds, instead of longer durations of over 700 or even 2000 seconds in EfficientKAN. HybridMLPEfficientKAN surpasses EfficientKAN in overall accuracy, exceeding 97% in BiT models. We also evaluate class-wise performance using recall, F1-score, specificity and Matthews Correlation-Coefficient (MCC). Hybrid approach effectively balances computational cost and prediction performance, making it an attractive solution for clinical settings where classification speed and accuracy are critical. This study highlights how BiT-based feature extraction combined with carefully designed models can provide efficient PBC recognition. The integration of MLP-level efficiency with KAN-style adaptability offers a promising avenue for developing high-accuracy, low-latency cell classification systems in hematological analysis.

This study presents a noncontact, in situ framework for etch depth prediction in plasma etching using machine learning (ML) and digital image colorimetry (DIC). While conventional ex situ methods offer accuracy, they suffer from delays and contamination risks. To overcome these, two approaches are explored. First, etch depth is initially obtained through ellipsometry mapping and used to train an artificial neural network (ANN) based on process parameters (e.g., plasma power, pressure, and gas flow), achieving significantly lower mean squared error (MSE) than a linear baseline. This is extended with a Bayesian neural network (BNN) to capture uncertainty in the predictions. Second, it is demonstrated that red, green, and blue data from DIC alone can effectively predict etch depth without relying on process parameters. Together, these findings establish ML-DIC integration as a real-time, low-cost, and noninvasive alternative for plasma process monitoring.