Search Results (13,205)

Background and Clinical Significance: Postpartum hemorrhage (PPH) is a leading cause of maternal mortality worldwide. Among its various etiologies, uterine atony accounts for approximately 70% of cases, while other causes include genital tract trauma, pathologic placentation, and intrapelvic arterial injury. Uterine artery embolization (UAE) has emerged as a preferred management option for severe PPH due to its high success rates of 89–98% and fertility preservation benefit. Despite its efficacy, UAE can lead to complications, such as pain, re-bleeding, infection, persistent vaginal discharge, ovarian insufficiency, and uterine necrosis—a rare but serious complication occurring in 1.4–2.7% of cases. Case Presentation: We present three cases of uterine necrosis following UAE from a single center (CHA Bundang Medical Center) between 2003 and 2024. All patients developed persistent high-grade fever approximately two weeks after the procedure, despite an initial response to antibiotic therapy. Imaging studies, including contrast-enhanced CT and MRI, revealed uterine ischemia and necrosis, and all patients ultimately required total hysterectomy. Conclusions: Uterine necrosis is a rare but potentially life-threatening complication of UAE that should be suspected in patients with persistent high-grade fever beyond the typical post-procedural course. Early imaging evaluation, particularly with contrast-enhanced modalities, is essential for prompt diagnosis. Timely surgical intervention, including hysterectomy, may be required to prevent severe morbidity. Full article

The theory of partially coherent image formation in dark-field and phase-contrast microscopy is presented. Explicit expressions and three-dimensional plots of the transmission cross-coefficient for different imaging modes are given. These include central and oblique internal dark field, annular dark field, phase contrast, differential phase contrast using semicircular or quadrant condenser pupils, and differential interference contrast. Explicit expressions are given for the image intensity for pure-phase objects consisting of a single-object spatial frequency or combinations of object frequencies. Full article

Stray-current corrosion from DC-electrified railways drives premature failure of buried metallic infrastructure (pipelines, foundations, tunnel reinforcement), causing resource waste, repair-driven carbon emissions and service disruptions that undermine the sustainability of urban transit corridors. Conventional impressed-current cathodic protection (ICCP) design relies on uniform-anode rules of thumb or closed commercial codes that cannot quantify the trade-off between protection uniformity, energy use and hardware cost. We present an open MATLAB framework that couples a custom 3D finite element method (FEM) solver with multi-objective particle swarm optimisation (MOPSO) and minimises three competing objectives simultaneously: total impressed current, RMS deviation from the protection target, and number of active anodes. A laboratory-calibrated coupling factor ($\mathrm{CF}=1.98$, consistent with the image-method prediction of 2 for a highly conductive pipe inclusion) absorbs the pipe–soil interface kinetics into a single direct FEM solve, and a pre-computed Green’s-function basis accelerates each MOPSO evaluation by more than two orders of magnitude. The solver is validated against an instrumented prototype with RMSE $=14.9$ mV across ten Cu/CuSO${}_4$ saturated reference electrode (CSE) measurements, and applied to a 500 m DC traction line. At an identical total current of 20.30 A across five anodes, the optimised design achieves an RMSE of 86.6 mV against the $-850$ mV NACE target, whereas a conventional uniform layout produces severe over-protection (RMSE $=1107$ mV)—a twelve-fold reduction. The framework is recommended as a transparent, reproducible engineering tool that simultaneously extends pipeline service life and reduces rectifier energy demand, supporting UN Sustainable Development Goals 9 and 11 for sustainable urban-rail infrastructure. Full article

Array Gm-APD LiDAR is highly vulnerable to strong backscattering caused by dynamic smoke. Conventional depth imaging methods cannot rapidly identify the smoke occlusion state, which greatly reduces the target recovery quality of the reconstructed depth image. To solve this problem, this paper presents a non-parametric algorithm for rapid smoke detection and depth imaging for array Gm-APD LiDAR. The proposed method does not rely on parameter estimation of the echo model. Instead, it determines the presence of smoke occlusion by calculating the Pearson correlation coefficient between the echo signal obtained from the superposition of all array pixels and the instrument response function. In this way, the method rapidly identifies smoke interference in a single depth image, performs fast denoising, and reconstructs the depth image. In a dynamic smoke environment with an average attenuation length of no more than 5.1, the proposed algorithm achieves 100% accuracy in occlusion discrimination based on 250 frames of array data. When the smoke occlusion rate reaches 96% and the average attenuation length is 2.29, the method obtains a target recovery of 0.71, which is 86.8% higher than that of the conventional algorithm. These results indicate that the proposed method has strong practical value for array Gm-APD LiDAR, especially for high-speed depth imaging in harsh atmospheric environments with severe obscuration. Full article

Background: Head trauma is a major public health concern. Computed tomography (CT) is frequently used to evaluate these patients but may expose them to unnecessary radiation exposure. Various biomarkers have been investigated to predict prognosis and reduce the need for unnecessary imaging. Red cell distribution width (RDW), neutrophil/lymphocyte ratio (NLR), and platelet/lymphocyte ratio (PLR) have been proposed as inflammatory markers; however, their diagnostic value in head trauma remains controversial. This study aimed to determine the value of complete blood count parameters in identifying intracranial injury in patients with mild head trauma. Methods: This prospective, single-center study enrolled 100 adults with mild head trauma. Demographic data, vital signs, neurological assessments, complete blood counts, CT results, and clinical outcomes were also recorded. Patients were categorized as intracranial injury positive (Group 1) or intracranial injury negative (Group 2). We statistically compared the laboratory and demographic data of the groups. Statistical significance was set at p < 0.05. Results: The study included 100 patients with mild head trauma who presented to the emergency department, of whom 11 were in Group 1. The median PLR and lymphocyte levels differed significantly between the groups (p < 0.05). Conclusions: The PLR may serve as a preliminary supportive marker to aid clinical assessment; however, its modest discriminatory performance suggests that it should not be used as a standalone diagnostic tool. Full article

Fine-grained insect pest classification presents a particularly demanding visual recognition challenge due to severe class imbalance, pronounced intra-class morphological variability across developmental stages, and high inter-class visual similarity among taxonomically related species. Existing deep learning approaches typically rely on a single feature representation extracted from a single network depth, overlooking complementary discriminative cues distributed across multiple abstraction levels. Furthermore, classical attention mechanisms perform spatial weighting deterministically, without explicitly modeling the underlying statistical structure of the feature space, which is inherently multimodal on long-tailed benchmarks such as IP102. This study proposes a Multi-Scale Gaussian Mixture Model-Gated Mixture of Experts (GMM-MoE) architecture that operates as a plug-in module insertable into any convolutional backbone, evaluated here on DenseNet-121 at three distinct feature depths. The proposed module computes analytic GMM posterior responsibilities in closed form, softly assigning each spatial location to dedicated convolutional expert sub-networks. At the same time, a conditional prior mechanism π(x) adapts the routing strategy to individual image content rather than relying on fixed priors. The architecture is evaluated on the IP102 benchmark (102 pest classes, ~75,000 images) under a two-stage training protocol. Ablation experiments confirm that increasing the number of experts consistently improves accuracy across all three routing depths, and that multi-scale fusion surpasses any single-scale configuration. The proposed model achieves a mean top-1 accuracy of 74.12% (±0.25%, 95% CI) across three independent runs on the IP102 test set. To the best of our knowledge, this is the first work to employ GMM posterior responsibilities as a spatial routing mechanism within a multi-scale CNN feature hierarchy for fine-grained insect pest classification, establishing a principled probabilistic alternative to deterministic attention weighting in visual recognition systems. Full article

This paper presents a column-parallel adaptive-gain single-slope (SS) analog-to-digital converter (ADC) for high-speed high-dynamic-range (HDR) CMOS image sensors. Conventional adaptive-gain approaches often rely on dual-ramp generation or duplicated column circuits, which increase area and power overhead. In contrast, the proposed architecture achieves adaptive-gain operation using a single global ramp shared across all columns. A reconfigurable capacitive attenuation network embedded inside each column comparator locally scales the ramp at the comparator input, enabling seamless transition between high-gain operation for low-level signals and unity-gain operation for large signals within a single exposure and readout cycle. To suppress mode-dependent offsets while maintaining low noise, a configurable dual-source-follower ramp buffer symmetrically buffers the ramp and reference voltages during auto-zeroing and is reconfigured as a full-sized buffer during unity-gain conversion. Switching-induced column offsets are compensated using optical black pixels and lightweight digital processing. The ADC is implemented in a 110 nm CMOS image sensor process and validated through post-layout simulations including extracted parasitics and Monte Carlo mismatch analysis. The core ADC consumes 36.8 µW per column. Simulation results demonstrate linearity error below 1% without missing codes and show that the proposed AG_x8-to-AG_x1 configuration extends the effective dynamic range up to 78.3 dB. Full article

Background/Objectives: Distinguishing primary lung squamous cell carcinoma (PLSCC) from metastatic head and neck squamous cell carcinoma (MHNSCC) to the lungs is challenging for pathologists, especially when patients present with a solitary lung nodule. The purpose of this study was to identify CT imaging features that differ between PLSCC from solitary MHNSCC to the lungs, using next-generation sequencing (NGS) and human papillomavirus in situ hybridization analysis as the gold reference standard. Methods: This retrospective, single-institution cross-sectional study included patients with a biopsy-proven PLSCC or solitary MHNSCC from July 2013 to May 2022, who underwent NGS or in situ hybridization, and baseline CT or PET/CT. Each scan was evaluated by at least two radiologists. Nodular, pleural, and ancillary CT features, as well as maximum standardized uptake value (SUVmax) from PET/CTs, were recorded. Associations between imaging features and pathology were examined using either the Wilcoxon rank-sum or Fisher’s exact test. Results: In total, 81 patients were included (median 66 years; 64 male); 36/81 (44%) had PLSCC and 45/81 (56%) had MHNSCC. PLSCC was associated with a larger size (median, 3.3–3.6 cm vs. 1.4–1.6 cm, p < 0.001), and the presence of post obstructive atelectasis (p = 0.002), pleural retraction (p < 0.001), pleural tags (p = 0.02), and pleural surface involvement (p = 0.02). MHNSCC presented as smaller peripheral nodules (p = 0.003) with lower SUVmax (p = 0.01). Conclusions: Several CT imaging features as well as SUVmax from PET/CT were significantly different between PLSCC and solitary MHNSCC and their potential discriminatory ability warrants evaluation in future studies. Full article

Deep learning-powered Street View Imagery (SVI) analytics provides a critical mechanism for smart city perception within the framework of Sustainable Development Goal 11 (SDG 11), effectively bridging the gap left by traditional remote sensing in fine-grained street-level observation. Over the years, deep learning-based semantic segmentation of urban streetscapes has become the dominant paradigm. However, when scaling to megacity measurements, current research faces the dual bottlenecks of “computational redundancy” and the “geographical domain shift” caused by the blind application of pre-trained models based on Western datasets. To address these challenges, this study is the first to systematically quantify the performance trade-off between Multi-Task Learning (MTL) and Single-Task Learning (STL) in megacity scenarios. Using this as a baseline, we constructed and validated a “low-computation, high-robustness” framework for streetscape semantic perception and spatial measurement. Relying on an integrated ResNeXt101-FPN MTL architecture and an ultra-low-cost fine-tuning strategy to overcome geographical domain shift, we extracted and analyzed the spatial heterogeneity of five core semantic elements—vegetation, sky, building, road, and vehicle—across the road network within Beijing’s 5th Ring Road. The results indicate the following: (1) We explicitly defined the computation-accuracy trade-off of MTL and STL in megacity perception. While utilizing only 1/5 of the parameters of STL, the MTL framework achieved a 5.34-fold increase in inference speed with a negligible 0.1% loss in overall mean Intersection over Union (mIoU); however, a 27.13% decrease in boundary segmentation accuracy was observed. (2) We established a low-cost, localized correction paradigm to overcome domain shift. Utilizing a minimal annotation cost (only 200 local images) significantly improved cross-domain adaptability, boosting the overall mIoU by 8.92% and significantly mitigating the geographical domain shift problem. (3) Multi-dimensional measurement and spatial analysis revealed a significant spatial decoupling pattern in Beijing’s streetscapes. The visual proportion of vegetation exhibited a pronounced “north-high, south-low” spatial differentiation, whereas built environment elements (e.g., building and road) displayed a typical “center-periphery” concentric gradient. This objectively reflects the spatial inequality of urban street greenery resources and the monocentric development characteristics of the built environment. The proposed framework therefore serves as a low-cost, AI-driven computational paradigm for smart city perception in resource-constrained regions. Furthermore, the revealed spatial heterogeneity offers data-driven insights for formulating sustainable urban renewal policies aligned with SDG 11. Full article

Background: Cardiac magnetic resonance (MRI) images often exhibit low contrast, anatomical variability, and indistinct boundaries, particularly in the myocardium (MYO) and right ventricle (RV). These challenges can reduce the reliability of both manual and automated segmentation, highlighting the need for more robust and boundary-aware approaches. Methods: In this study, an Aspect-Aware Boundary-Resilient UNet3D (ABR-UNet3D) architecture is proposed for cardiac MRI segmentation. The model incorporates an Aspect-Aware Complementary Attention (AAC) module that combines multi-planar contextual information with a complementary gating mechanism to enhance boundary representation. The method was evaluated on the ACDC dataset under consistent training conditions. In addition to Dice Similarity Coefficient (DSC) and Intersection over Union (IoU), boundary-based metrics, including the 95th percentile Hausdorff Distance (HD95), Average Surface Distance (ASD), and Surface Dice, were employed. Furthermore, a five-fold cross-validation protocol and detailed ablation studies were conducted to assess robustness and analyze the contribution of individual AAC components. Results: The proposed method achieved a mean DSC of 0.9603 in single-run experiments on the ACDC dataset and showed consistent performance in anatomically challenging regions, particularly for RV and MYO segmentation. In addition, five-fold cross-validation experiments resulted in an average DSC of 0.952 ± 0.009 and IoU of 0.908 ± 0.012, indicating stable performance across different data splits within the evaluated dataset. Boundary-based metrics also showed improved surface agreement and lower boundary errors compared with the evaluated baseline models. Ablation studies further indicated that the combined use of multi-planar contextual information and complementary gating contributes more effectively to segmentation performance than the individual components used separately. Conclusions: The results suggest that the proposed ABR-UNet3D architecture provides a stable and competitive segmentation framework for cardiac MRI images within the scope of the ACDC dataset. By jointly modeling contextual information and boundary refinement, the method improves segmentation reliability in challenging regions while maintaining competitive and consistent performance with respect to existing approaches. Full article

To address the challenge of accurately reconstructing multi-depth targets using single-photon Light Detection and Ranging (LiDAR) under few-frame conditions in high-precision applications such as autonomous driving perception, remote sensing, and military reconnaissance, this paper proposes a computational imaging method named the Jensen–Shannon Divergence Weighted Pixel Fusion Constant False Alarm Rate (JSWPF-CFAR) approach. First, the proposed method utilizes the Jensen–Shannon (JS) divergence to characterize the statistical similarity between adjacent pixels, thereby constructing adaptive weights to achieve the effective fusion of echo signals. The key innovation lies in the formulation of a JS divergence-based weighting factor, which fully exploits the inherent spatial correlation within 3D target structures to optimize the pixel fusion process and enhance the signal statistics of target echoes. Subsequently, a CFAR detection model tailored for Geiger-mode Avalanche Photodiode (GM-APD) multi-depth echo signals is constructed to estimate the noise photon count within a local sliding window; this estimate is then used to calculate a photon counting threshold for identifying and extracting high-confidence target intervals. Finally, a peak-picking method is employed to perform the 3D reconstruction of multi-depth targets. Compared with existing techniques such as matched filtering and Reversible Jump Markov Chain Monte Carlo (RJMCMC), the proposed method exhibits superior reconstruction quality under few-frame and low Signal-to-Background Ratio (SBR) conditions. The experimental results demonstrate that the proposed method achieves an improvement in target restoration degree (RD) of at least 21.16% and a relative variance (Var) optimization of at least 62.90% over the matched filtering and RJMCMC baselines. These results indicate that the proposed approach effectively enhances the multi-depth estimation performance of single-photon LiDAR in complex scenes. Full article

Digital camera sensors are designed to capture a wide range of incident illuminants, enabling the creation of high-quality images. However, these sensors lack the capability to differentiate between the color of the source illuminant and the actual color (or original color) of the object being captured. For this reason, the computational color constancy (CCC) was introduced and has been developed over decades. The CCC is an approach to modeling the color perception of the human visual system (HVS) by ensuring accurate object color determination under varying source illuminant conditions. At the core of human visual perception (HVP)-based CCC is attaining higher accuracy in scene illuminant estimation. The emergence of deep convolutional neural networks (DCNNs) was a recent innovation in accurate illuminant estimation, fundamentally transforming the CCC research landscape. Nevertheless, accurate illuminant estimation still remains a huge challenge for both traditional and state-of-the-art (SOTA) approaches. To further advance precision in illuminant estimation, this article presents a novel learning-based illumination color cast estimation approach to HVP-based CCC. Most importantly, the proposed approach is intended to integrate informative features into both channel and spatial regions while preserving long-term dependency feature information with the use of dense skip connections. To achieve these objectives, the proposed Dense Dual Connection Aggregated Transform Network (DDCATNet) architecture is designed to comprise several modules: shallow feature extraction, channel-wise and spatial feature-based Dense Dual Connection (DDC), fusion of the dense channel-wise attention (CA) and spatial attention (SA) branches through a gate mechanism (GM) unit, and aggregate transform. It is worth noting that both the CA blocks and the SA blocks in the DDC module are characterized by dense and cascading connections, meant to preserve long-term feature information and modulate different-level feature information at both global and local scales. The densely connected CA branch (DCA) and the densely connected SA branch (DSA) are also highly effective in securing high-contribution information while suppressing redundant data. The GM unit is integrated at the back of the DDC module, fusing the two DCA and DSA branches to ensure the adaptive merging of useful hierarchical feature information and the extraction of more valuable feature information. As a result, the proposed DDCATNet architecture significantly enhanced precision in illuminant estimation, thereby improving performance. In rigorous experiments on a wide range of datasets, the proposed DDCATNet approach outperformed its SOTA counterparts, validating the efficacy and generalization capabilities, as well as robust camera-invariance, across diverse, single- and multi-illuminant datasets and model architectures. Full article

Objective: To provide a comprehensive literature review of original work on artificial intelligence in ocular oncology. Methods: Scoping review of PubMed-indexed original articles (n = 94) on the use of artificial intelligence in ocular oncology, retrieved during the month of February 2026 and independently screened by two ocular oncologists. Results: Most of the literature on artificial intelligence (AI) in ocular oncology focuses on uveal melanoma and its differentials (n = 39, 41%), followed by retinoblastoma (n = 14, 15%) and orbital tumors (n = 12, 13%). The purpose of using the AI models was to screen, diagnose, and classify the disease (n = 59, 62%) and to treat, predict outcomes, and monitor the disease (n = 35, 37%). Most literature (n = 32, 34%) on AI in ocular oncology originates from China. Datasets comprised images in 78% (n = 73) of the studies, clinical parameters in 14% (n = 13), and omics data in 12% (n = 11). Most studies worked on developing AI models (n = 83, 88%), of which two reached a deployment stage. Few studies evaluated or incorporated pre-existing models (n = 11, 12%). Supervised learning strategy was most commonly employed (n = 75, 80%). Among studies that developed AI models, traditional machine learning architectures were used in 36, deep learning in 39, and a combination in 8. Most studies (n = 59, 63%) were at a Clinical AI Readiness Evaluator Technology Readiness Level 4, i.e., at the prototype development stage. Conclusions: Despite the limitation of a single database search, a surge in AI applications in ocular oncology after 2020 is evident. Most studies are in the model development stage, and few have been deployed in the real world for clinical implementation. Very few models have proven effective in real-world clinics and the community, holding promise for the future. Full article

Background: Tirbanibulin 1% ointment has demonstrated short-term efficacy and excellent tolerability in the treatment of actinic keratosis (AK) on the face and scalp. However, data on long-term efficacy are still lacking. Materials and Methods: This prospective, single-center, 12-month extension study included patients with facial and scalp AKs previously treated with tirbanibulin 1% ointment once daily for 5 consecutive days. Long-term analysis was restricted to lesions that had achieved complete clinical and dermoscopic clearance at the 2-month follow-up. At 12 months, the treated areas were reassessed clinically and dermoscopically. High-resolution images obtained at baseline, 2 months, and 12 months were compared lesion by lesion to distinguish sustained clearance, recurrence at the same anatomical site, and the development of new AKs within the treated field. Results: Thirty-seven patients were reassessed at 12 months. Of the 228 AKs treated at baseline, 116 lesions had achieved complete clearance at 2 months and were therefore eligible for long-term evaluation. At 1 year, 70/116 lesions (60.3%) remained free of recurrence, whereas 46/116 (39.7%) relapsed. Sustained clearance was observed in 35/51 grade 1 lesions (68.6%), 32/57 grade 2 lesions (56.1%), and 3/8 grade 3 lesions (37.5%). In addition, 35 new AKs developed within the previously treated field. No delayed local or systemic adverse events and no progression to invasive cSCC were observed during follow-up. Patient-reported satisfaction was high, and 94% of patients stated they would be willing to repeat the treatment. Conclusions: Tirbanibulin was associated with sustained lesion clearance at one year, particularly in lower-grade AKs. While recurrence remains relatively common—especially in thicker lesions—the treatment was well tolerated and associated with no delayed adverse effects. Its short application regimen and excellent safety profile support tirbanibulin’s role in the long-term management of field cancerization. Full article

Background: Hemorrhagic transformation is a major complication after thrombolysis in acute ischemic stroke, and its risk assessment based on baseline non-contrast CT (NCCT) remains limited. This study aims to evaluate the role of CT-based texture analysis of the infarcted region on baseline NCCT in characterizing tissue vulnerability to hemorrhagic transformation. Methods: This retrospective single-center study included 50 patients with supratentorial acute ischemic stroke treated with thrombolysis. All patients underwent baseline NCCT and follow-up CT at 24 h and were divided into those with hemorrhagic transformation (n = 33) and those without (n = 17). Texture analysis was performed using MaZda software, and feature selection was conducted using the Fisher coefficient. Discriminatory ability was assessed using receiver operating characteristic analysis. Results: Gray-level and run-length non-uniformity features were significantly lower in patients with hemorrhagic transformation (p < 0.001). These parameters yielded AUCs of 0.94–0.97, highlighting their potential as exploratory imaging biomarkers. Conclusions: CT-based texture analysis of baseline NCCT may provide additional quantitative information associated with subsequent hemorrhagic transformation. However, given the small sample size, single-center design and absence of external validation, these findings should be interpreted as exploratory and hypothesis-generating. Full article