Search Results (536)

Brain tumor detection from magnetic resonance imaging (MRI) is fundamental to computer-aided diagnosis, yet automated models must remain robust to heterogeneous imaging conditions. Despite strong recent progress, many deep learning and transformer-based approaches primarily optimize performance accuracy without explicitly improving feature selectivity and spatial localization, and they typically produce deterministic output without uncertainty estimates, which limits reliability. To overcome these limitations, we introduce UA-EffNet-DA, an uncertainty-aware EfficientNet framework that addresses these limitations through three complementary components. First, EfficientNet-B4 serves as an efficient backbone for discriminative feature extraction. Second, lightweight dual attention modules, comprising channel and spatial attention in parallel, are applied to the model to emphasize what and where discriminative features to focus within MRI slices. Third, Monte Carlo dropout is employed during inference to quantify predictive uncertainty and enable confidence-aware decision. Experiments on two public benchmarks demonstrate strong performance, yielding accuracies of 98.73% on the Figshare dataset and 99.23% on the Kaggle dataset. In addition, explainable AI analysis using Gradient-weighted Class Activation Mapping (Grad-CAM) further indicates that the proposed model concentrates on diagnostically relevant tumor regions rather than background structures, supporting transparent decision-making. Ablation studies confirm the complementary contribution of dual attention refinement and uncertainty-aware inference. Overall, the proposed UA-EffNet-DA framework offers an efficient and interpretable approach for brain tumor detection that supports more reliable clinical decision support through uncertainty-aware predictions. Full article

Accurate brain tumor segmentation in magnetic resonance imaging (MRI) is essential for diagnosis, treatment planning, and therapy monitoring. Conventional deep learning models often struggle with large variations in tumor shape, size, and contrast, as well as severe foreground–background imbalance. To address these challenges, this study presents Deep-MiSR, an enhanced encoder–decoder framework built upon DeepLabV3+ with a MobileNetV2 backbone, tailored for single-modality contrast-enhanced T1-weighted (T1CE) MRI segmentation. Three complementary components are integrated into the architecture: mixed depthwise convolution (MixConv) with heterogeneous kernels within the atrous spatial pyramid pooling module for multi-scale feature aggregation, a squeeze-and-excitation block for adaptive channel recalibration, and R-Drop regularization that enforces prediction consistency via symmetric Kullback–Leibler divergence. The model was evaluated on 3064 T1CE slices from 233 patients drawn from the publicly available Nanfang Hospital brain MRI dataset. Deep-MiSR achieved a Dice similarity coefficient of 0.9281, a mean intersection-over-union of 0.8738, a precision of 0.8839, and a 95th-percentile Hausdorff distance of 7.69 mm, demonstrating consistent improvements over both the DeepLabV3+ baseline and all prior methods evaluated on the same data. Ablation studies confirmed that each component contributes independently, with R-Drop providing the largest individual gain. These findings demonstrate that combining multi-scale convolution, channel attention, and consistency regularization constitutes an effective and computationally practical strategy for robust single-modality brain tumor segmentation. Full article

Background/Objective: Cardiac innervation plays a critical role in regulating myocardial function and enabling the heart to adapt to physiological and pathological conditions. Although the general features of sympathetic and parasympathetic innervation of the myocardium are well described, the spatial organisation of nerve fibres within the cardiac muscle remains incompletely characterised. This study aimed to develop and validate the SKUF (Slice–Keep–Unwrap–Fuse) protocol, a multimodal framework for mapping myocardial innervation through the integration of histological data and magnetic resonance imaging (MRI). Methods: The study was performed on the heart of a 7-year-old patient who died from rupture of a cerebral vascular malformation without evidence of cardiovascular disease. Prior to histological processing, post-mortem MRI was performed to provide a precise anatomical reference. The heart was sectioned into sequential transverse rings of 4 mm thickness, yielding 71 paraffin blocks. Histological sections (3 μm) were immunostained with antibodies against UCHL-1 to visualise nerve fibres and scanned using an Aperio AT2 system (20× magnification). Automated image analysis was conducted using the SVSSlide Processor module, which included tissue segmentation, colour-based nerve fibre detection, and sliding-window density mapping. Heatmaps were assembled into ring-based myocardial reconstructions and co-registered with MRI slices using combined rigid and deformable registration, followed by three-dimensional reconstruction of innervation patterns. Results: A higher density of nerve fibres was observed in the right ventricular myocardium compared with the left ventricle, whereas larger nerve trunks were identified in the epicardium of the left ventricle. Quantitative analysis revealed a pronounced longitudinal gradient of innervation, with minimal density in the apical region and progressive increases towards the mid-ventricular segments, where maximal density and spatial organisation of neural structures were observed. The atrioventricular groove exhibited the greatest heterogeneity of innervation due to the presence of large nerve trunks and ganglionated plexuses. Integration of histological maps with MRI enabled three-dimensional visualisation of spatial clusters of nerve fibres. Conclusions: The SKUF protocol provides a robust framework for integrating histological and MRI data to generate three-dimensional maps of myocardial innervation. This approach may facilitate the development of high-resolution anatomical atlases of cardiac innervation and support future studies of neurocardiac mechanisms of arrhythmogenesis and targeted neuromodulation. Full article

Translational neuroscience relies on both in vitro slice recordings and in vivo recordings. Their spontaneous population dynamics are observed under decisively different conditions, and across independent experiments, there is typically no clear neuron-to-neuron correspondence. Here, we formulate a one-step-ahead, 1 ms binned, bidirectional transfer task between in vitro and in vivo multineuronal spike trains and provide a standardized evaluation procedure for generation across markedly different recording preparations. We train an autoregressive transformer on 1 ms binned, 128-unit binary sequences and introduce Dice loss to directly optimize spike-event overlap under extreme class imbalance, comparing it with Binary Focal Cross-Entropy (γ = 2.0). Across 12 mouse datasets (6 in vitro HD-MEA sessions and 6 in vivo Neuropixels sessions), the method achieves strong within-domain performance and remains above chance for cross-domain generation (ROC-AUC 0.70 ± 0.09 for in vitro → in vivo; 0.80 ± 0.10 for in vivo → in vitro). Because spike events are rare, we report Precision–Recall curves and PR-AUC alongside ROC-AUC to reflect minority-event quality. The present results should be interpreted as predictive generation under preparation/domain shift rather than as direct evidence of preserved causal biological dynamics; whether the framework also reflects features such as E/I balance or oscillatory structure remains an important question for future validation. To our knowledge, this is the first demonstration of bidirectional, time-resolved generation between unpaired in vitro and in vivo population spike trains without assuming cell correspondence, and the framework can be adapted to other sparse neural event data and related event-based datasets when domain-specific validation criteria are defined. Full article

Background/Objectives: Acute pancreatitis (AP) is a significant inflammatory pancreatic disease with high morbidity and mortality rates that requires early and accurate diagnosis. In this study, a deep learning-based classification system is developed and evaluated for the automatic classification of AP and normal pancreas from contrast-enhanced CT images, with a focus on patient-level assessment to enhance clinical applicability. Methods: A study-specific dataset is created for the study in CT images from 183 patients (103 normal and 80 with AP). To prevent data leakage and objectively evaluate model performance, the dataset is divided into training and test sets based on patient-level data. Convolutional neural network (CNN)-based architectures, such as ResNet50, EfficientNet, and ConvNeXtV2, are compared with Transformer-based architectures, such as Swin Transformer (Swin) and Vision Transformer (ViT). Results: In slice-level analysis, all models achieve high performance. Swin shows the highest accuracy (84.06%), and ViT revealed the most balanced performance with an F1-score of 82.90%. In the more clinically significant patient-level evaluation, the ViT model outperforms all others with an accuracy of 89.19%, an F1-score of 86.67%, and an area under the curve (AUC) of 0.946. The ViT model’s high AUC and recall values demonstrate its ability to reliably distinguish between AP and normal pancreas classes, even under different threshold values. Conclusions: These results suggest that transformer-based architectures can extract stronger and more reliable feature representations from pancreatic CT images due to their capacity to model global contextual features. Furthermore, the patient-level evaluation approach enables the model to generate results that are more compatible with clinical decision-making processes, thereby enhancing its usability in real clinical settings. In conclusion, the proposed ViT-based approach is a promising method for diagnosing acute pancreatitis. Full article

Objective: This study investigates the influence of selecting three thermoplastics as raw materials (PLA, PETG, and ABS) on dimensional accuracy, defect formation, and slicer-based mass prediction reliability in FFF 3D-printed UAV micro-frames. Methods: A factorial experimental design combining three materials, two micro-frame geometries, and two infill levels was implemented. Print quality was assessed through structured visual inspection of common FFF defects, while manufacturing reliability was evaluated by comparing slicer-predicted and experimentally measured mass. Dimensional fidelity was quantified at critical motor mount features using repeated micrometric measurements and dedicated accuracy and uniformity indices. Results: The results reveal strong material-dependent behaviour. PLA exhibited the highest dimensional consistency and near-zero mean mass prediction error, PETG showed intermediate performance, and ABS presented significant warping, together with a pronounced positive mass prediction bias. These findings indicate systematic discrepancies between predicted and measured mass values and highlight the need for material-dependent calibration of slicing software. Conclusions: Material selection and process calibration strongly affect dimensional fidelity and manufacturing reliability in FFF-printed UAV micro-frames. The findings provide practical guidance for material choice and slicing parameter adjustment in UAV fabrication and similar small-scale FFF applications. Full article

Hyperspectral remote sensing provides essential biochemical and structural information for crop disease monitoring, yet its application to cotton Verticillium wilt has largely focused on single-period evaluations or multi-temporal classifications. Such approaches overlook the progressive nature of this vascular disease, whose pigment, water, and mesophyll responses evolve over time, making temporal hyperspectral information critical for reliable severity estimation but still insufficiently utilized. To overcome this limitation, we conducted daily time-series observations on cotton leaves and collected 2895 hyperspectral reflectance measurements and 770 high-resolution RGB images together with disease severity records, generating a temporally dense spectral-severity dataset spanning symptom-free to severe stages. Five categories of disease-related vegetation indices were derived and organized into 5-day spectral–temporal slices. Based on these features, we introduce a dual-branch Transformer-TCN model that integrates global temporal dependencies captured by self-attention with local temporal variations resolved by dilated causal convolutions for severity inversion. The model delivers the strongest performance with an R² of 0.8813, exceeding multiple single and hybrid time-series alternatives by 0.0446–0.1407 in R², equivalent to a relative improvement of 5.33–19.00%. Temporal spectral features also outperform their non-temporal counterparts, highlighting that disease progression dynamics captured by time-series spectra are critical for reliable severity retrieval. Feature contribution analysis indicates that the blue red index BRI provides the highest contribution, consistent with the single-index time-series modelling results. Photosynthesis- and water-related indices provide secondary but complementary support. Collectively, our results demonstrate that the dual-branch Transformer-TCN model can capture complex spectral–temporal relationships between cotton Verticillium wilt and disease severity, providing methodological support for crop disease monitoring and evaluation. Full article

Radiation pneumonitis (RP), one of the most common and severe complications in locally advanced non-small cell lung cancer (LA-NSCLC) patients following thoracic radiotherapy, presents significant challenges in prediction due to the complexity of clinical risk factors, incomplete multimodal data, and unavailable slice-level annotations in pre-radiotherapy CT images. To address these challenges, we propose a multimodal fusion framework based on Dual-Layer Attention-Based Adaptive Bag Embedding Multiple-Instance Learning (DAAE-MIL) for accurate RP prediction. This study retrospectively collected data from 995 LA-NSCLC patients who received thoracic radiotherapy between November 2018 and April 2025. After screening, Subject datasets (n = 670) were allocated for training (n = 535), and the remaining samples (n = 135) were reserved for an independent test set. The proposed framework first extracts pre-radiotherapy CT image features using a fine-tuned C3D network, followed by the DAAE-MIL module to screen critical instances and generate bag-level representations, thereby enhancing the accuracy of deep feature extraction. Subsequently, clinical data, radiomics features, and CT-derived deep features are integrated to construct a multimodal prediction model. The proposed model demonstrates promising RP prediction performance across multiple evaluation metrics, outperforming both state-of-the-art and unimodal RP prediction approaches. On the test set, it achieves an accuracy (ACC) of 0.93 and an area under the curve (AUC) of 0.97. This study validates that the proposed method effectively addresses the limitations of single-modal prediction and the unknown key features in pre-radiotherapy CT images while providing significant clinical value for RP risk assessment. Full article

This study explored the feasibility of near-infrared (NIR) spectroscopy for detecting total water, free water and bound water in carrot slices during freeze-drying, with low-field nuclear magnetic resonance (LF-NMR) characterizing water state distribution and oven-drying determining moisture content (MC). NIR spectra (10,000–4000 cm⁻¹) were processed via optimized sample partitioning, preprocessing and feature extraction; partial least squares regression (PLSR), support vector regression (SVR), back-propagation artificial neural network (BPANN), extreme gradient boosting (XGBoost) and particle swarm optimization–random forest (PSO-RF) models were established and evaluated. Results showed that SVR and BPANN performed robustly, with CARS being the optimal feature extraction method. The full-moisture system achieved high total/free water prediction accuracy ($R_p^2$ = 0.9902/0.9740), while the low-moisture system improved bound water prediction ($R_p^2$ = 0.9709). The established NIR models exhibited excellent fitting and generalization ability, enabling rapid and non-destructive quantitative prediction of moisture content during carrot freeze-drying. Full article

The rapid growth of 5G necessitates wireless receivers capable of high-speed, low-latency communication under complex channel conditions. Traditional receivers struggle with the performance–complexity trade-off in massive MIMO systems, where linear detectors underperform and maximum likelihood (ML) detection becomes computationally prohibitive. Deep-learning-based model-driven approaches have demonstrated a favorable balance between detection performance and computational cost. However, despite their algorithmic promise, the transition of these learned detectors into practical, real-time systems is critically hampered by inefficient hardware mapping, resulting in suboptimal throughput, high resource overhead, and limited scalability. To bridge this gap, this paper presents LwAMP-Net, a dedicated FPGA accelerator for a lightweight learned AMP detector. We propose a modular and multi-mode hardware architecture for LwAMP-Net, featuring an outer-product-based dataflow that mitigates pipeline stalls and multi-mode processing elements that adapt to diverse computation patterns. These innovations jointly enhance computational parallelism and resource utilization on the FPGA. Implemented on a Xilinx XC7VX690T FPGA for a 128 × 8 MIMO system with 16QAM, the accelerator achieves a 49.2% higher normalized throughput per iteration, an 85.4% improvement in throughput per LUT slice, and a 12.7% improvement in throughput per DSP compared to the state-of-the-art methods. This work provides a complete architectural solution for deploying high-performance, hardware-efficient learned MIMO detectors in real-world systems. Full article

To address over-extrusion and forming defects at path corners caused by path overlap in additive manufacturing, this paper proposes an angle-adaptive look-ahead compensation algorithm based on a Sech attenuation curve. This method establishes a mapping model between the path angle and the adaptive look-ahead distance of the overlapping area, aiming to eliminate the material accumulation at the corner by precisely identifying the overlapping area and modulating the flow rate. By building a Beckhoff five-axis 3D-printing device and relying on the TwinCAT control platform, the compensation triggering logic based on PLC real-time Euclidean distance calculation was realized, and a slicing software with dynamic bias compensation was also developed. Experiments were conducted on triangular samples with extreme acute angles of 5°, universal acute angles of 85°, and 90° straight angles for printing verification. The results show that this algorithm can effectively suppress the material over-extrusion and accumulation at the path overlap in multiple angles, achieving a smooth transition of the sharp corners in the printed contour. The research confirms that the algorithm proposed in this study, together with the integrated software and hardware system, can ensure the forming accuracy of extreme and conventional geometric features while also guaranteeing the printing efficiency, providing an important reference for ensuring the quality coordination control of the formation process of extreme geometric features in additive manufacturing. Full article

Background/Objectives: This study aimed to examine the prevalence of anatomical variations in the sacroiliac joints (SIJs) as observed through Magnetic Resonance Imaging (MRI), to characterize their manifestations, and to identify MRI features that may resemble inflammatory alterations. Methods: A retrospective review was conducted on consecutive MRI scans of the SIJ performed from January 2009 to January 2022. Eight anatomical variations, along with associated edematous and structural changes, were assessed. Results: The study encompassed 840 patients, with anatomical variations identified in 39.7% of the cohort, occurring more frequently among female participants. The most prevalent variations were accessory SIJ (36.2%) and the iliosacral complex (32.2%). Notably, isolated synostosis and persistent ossification center variations were absent. The increased frequency of variations in women, as well as their correlation with advancing age, was statistically significant (p = 0.034). Accessory SIJ and dysmorphic alterations were linked to bone marrow edema and structural modifications. In the iliosacral complex and semicircular defect variations, prominent vascular structures were observed extending along the bone surfaces. The number and depth of edema slices in sacroiliitis exceeded those observed in the variation (p < 0.001). Conclusions: Anatomical variations of the SIJ are prevalent among women and tend to increase with advancing age. Given that these variations, particularly accessory SIJ and dysmorphic alterations, may present with edematous and structural signal intensity changes that resemble sacroiliitis, it is crucial to recognize these variations. It is recommended to assess axial and coronal images concurrently and to exercise caution in the interpretation of SIJ MR images. Full article

Computed tomography (CT) enables non-destructive insight into internal log structure, yet fully automated interpretation of CT images remains limited by inconsistent annotations, boundary ambiguity, and insufficient spatial context in 2D slice-based analysis. These challenges restrict the industrial deployment of deep learning for wood quality assessment. This study applies artificial intelligence (AI) and deep learning to the automated analysis of computed tomography (CT) scans of wood logs for detecting internal qualitative features and segmenting bark. Using convolutional neural networks (CNNs), trained models accurately distinguish healthy and damaged regions and segment bark, including discontinuous parts. We introduce a novel pseudo-spatial representation by merging consecutive slices into red–green–blue (RGB) format, which improves prediction accuracy and model robustness across logs. To enhance interpretability, Gradient-weighted Class Activation Mapping (Grad-CAM) highlights regions contributing most to defect detection, particularly knots. Comprehensive evaluation using Sørensen–Dice similarity coefficients and confusion matrices confirms the effectiveness of the proposed approach under industrial conditions. These findings demonstrate that AI-driven CT image analysis can address key limitations of current log-grading workflows and enable more reliable, objective, and scalable quality assessment for timber-dependent economies. Full article

Despite significant progress in general-purpose diffusion-based models capable of producing high-quality media, this approach is still too difficult to implement on consumer/gamer hardware. We present here a memory-optimized DreamBooth framework designed for consumer-grade GPUs with 16 GB of VRAM, that allows for end-to-end image personalization and addresses some of the limitations of existing solutions. Our system reduces peak GPU memory from 22 GB (baseline DreamBooth) to 14.2 GB through novel hierarchical memory management, including attention slicing, Variational Autoencoder (VAE) tiling, gradient accumulation, and gradient checkpointing integrated within the Hugging Face Accelerate ecosystem. The framework further incorporates state-of-the-art techniques for preserving facial features and a comprehensive automated quality management system. The result is a complete end-to-end pipeline achieving a peak memory of 14.2 GB, with quantitative performance (LPIPS: 0.139, SSIM: 0.879, identity: 0.852, and FID: 23.1) competitive with methods requiring significantly more hardware resources. Full article

Under high renewable energy penetration, wind power forecasting faces pronounced challenges due to strong randomness and uncertainty, making conventional point-forecast-centric paradigms insufficient for risk-aware and reliable power system scheduling. An uncertainty-aware temporal transformer framework for wind power forecasting is presented, integrating probabilistic modeling with deep temporal representation learning to jointly optimize prediction accuracy and uncertainty characterization. Crucially, rather than treating uncertainty quantification merely as a post-processing step, the central conceptual contribution lies in modularizing uncertainty directly within the attention mechanism. A probability-driven temporal attention mechanism is incorporated at the encoding stage to emphasize high-variability and high-risk time slices during feature aggregation, while a multi-quantile output and interval modeling strategy is adopted at the prediction stage to directly learn the conditional distribution of wind power, enabling simultaneous point and interval forecasts with statistical confidence. Extensive experiments on multiple public wind power datasets demonstrate that the proposed method consistently outperforms traditional statistical models, deep temporal models, and deterministic transformers, as validated by formal statistical significance testing. Specifically, the method achieves an MAE of $0.089$, an RMSE of $0.132$, and a MAPE of $10.84\%$ on the test set, corresponding to reductions of approximately $8\%$–$10\%$ relative to the deterministic transformer. In uncertainty evaluation, a PICP of $0.91$ is attained while compressing the MPIW to $0.221$ and reducing the CWC to $0.241$, indicating a favorable balance between coverage reliability and interval compactness. Compared with mainstream probabilistic forecasting methods, the model further reduces RMSE while maintaining coverage levels close to the 90% target, effectively mitigating excessive interval conservatism. Moreover, by adaptively generating heteroscedastic intervals that widen during high-volatility events and narrow under stable conditions, the model achieves a highly focused and effective capture of critical uncertainty information. Full article