Fracture detection in tunnel borehole images using MSFDNet: An interpretable multi-scale network with partial convolution and bi-level routing attention

An intelligent measurement framework for multi-type rock detection and classification in geological images

Softmatch distance: A novel distance for weakly-supervised trend change detection in bi-temporal images

Ensuring reliable digital pathology: a comparative analysis of HistoQC and PathProfiler for artefacts detection in prostate whole-slide images.

Optimized multi-dimensional attention spiking neural network for pneumonia detection in chest x-ray images

Dual-path AI-generated image detection: Leveraging texture-rich and texture-poor patches with global semantic features

Hierarchical mobile-dense convolutional architecture for tampered image detection using focal optimization with quantized edge TPU deployment

SFNet: Hierarchical perception and adaptive test-time training for AI-generated military image detection

The concrete cracking problem can seriously affect the durability and safety of civil structures. Accurately and quickly measuring the width of concrete cracks can help control defect development in a timely manner. Current research mainly relies on pixel detection of two-dimensional images, which lacks real three-dimensional information about crack lesions. Detection results are also obviously affected by various factors, such as shooting distance and posture, resulting in poor accuracy. Therefore, this paper presents an engineering-integrated solution that combines U-Net-based crack segmentation with binocular vision 3D reconstruction. The focus is placed on the practical deployment of the integrated pipeline, the optimization of key parameters under real inspection conditions, and the experimental validation of measurement accuracy on actual concrete cracks. Firstly, the U-Net deep learning algorithm is used to automatically identify and segment the concrete crack region; then, a binocular vision-based 3D reconstruction pipeline is adopted, and a parallax rejection algorithm based on a “double-threshold” decision is proposed to improve the fidelity of crack disparity maps, and the effect of the filter window size on the concrete crack region is analyzed; finally, an intelligent measurement method based on the 3D reconstruction model is proposed, and the measurement results of concrete crack width can be calculated directly from the 3D reconstruction model. The results show that (1) the model can identify the characteristics of the crack, and the detection effect at 4:00 p.m. is the best, because at this time the light is more uniform with less shadow and moderate contrast between the crack and its background; (2) the reconstruction of the 3D point cloud model of the concrete crack with a filtering window of size 9 × 9 is the best; (3) the maximum error between the calculated and measured values of crack width is 0.31mm, the minimum error is 0.07mm, and the average error is 0.15 mm, which indicates that the measurement accuracy reaches the sub-millimetre level and verifies the validity of the proposed method in this paper.

Retinal Diabetic Macular Edema Detection in OCT Images Using DeepLabv3 + with ResNet-18 Architecture

Developing YOLOv5 for weld flaws detection and localization in gamma radiography images based on attention mechanisms

Up-to-date and reliable detection of oil spills is important to protect marine ecosystems and allow fast and accurate responses. In this approach, a full deep learning system for automatic oil spill detection and classification in aerial RGB images acquired from UAVs is presented. A dual- attention semantic segmentation network is selected by the system to improve the feature extraction for images taken in challenging marine ecosystems, and also a GAN-based data augmentation approach to reduce the issue of sparsely annotated data. With clear differences in appearance, using the proposed method the distinct visually identifiable oil types: rainbow, silver, brown, and black oils can all be identified and distinguished. The method also allows the generation of segmented spill maps for area and volume estimation. Experiments show that the system consistently achieves higher segmentation accuracy than traditional models. It also provides a solution to scalable real- time and practical monitoring of marine environments.

Correction: A complete benchmark for polyp detection, segmentation and classification in colonoscopy images

Background/Objectives: Early and accurate detection of breast cancer is essential for reducing mortality and improving patient outcomes. However, the manual interpretation of breast ultrasound images is challenging due to image variability, noise, and inter-observer subjectivity. This study aims to address these limitations by developing an automated and interpretable computer-aided diagnosis (CAD) system. Methods: We propose an automated and interpretable computer-aided diagnosis (CAD) system that integrates ensemble transfer learning with Vision Transformer architectures. The system combines the Data-Efficient Image Transformer (Deit) and Vision Transformer (ViT) through concatenation-based feature fusion to exploit their complementary representations. Preprocessing, normalization, and targeted data augmentation enhance robustness, while Gradient-weighted Class Activation Mapping (Grad-CAM) provides visual explanations to support clinical interpretability. The proposed model is benchmarked against state-of-the-art CNNs (VGG16, ResNet50, DenseNet201) and Transformer models (ViT, DeiT, Swin, Beit) using the Breast Ultrasound Images (BUSI) dataset. Results: The ensemble achieved 96.92% accuracy and 97.10% AUC for binary classification, and 94.27% accuracy with 94.81% AUC for three-class classification. External validation on independent datasets demonstrated strong generalizability, with 87.76%/88.07% accuracy/AUC on BrEaST, 86.77%/85.90% on BUS-BRA, and 86.99%/86.99% on BUSI_WHU. Performance decreased for fine-grained BI-RADS classification-76.68%/84.59% accuracy/AUC on BUS-BRA and 68.75%/81.10% on BrEaST-reflecting the inherent complexity and subjectivity of clinical subclassification. Conclusions: The proposed Vision Transformer-based ensemble demonstrates high diagnostic accuracy, strong cross-dataset generalization, and clinically meaningful explainability. These findings highlight its potential as a reliable second-opinion CAD tool for breast cancer diagnosis, particularly in resource-limited clinical environments.

TrustNet: a lightweight network with integrated uncertainty quantification and quantitative explainable AI for ischemic stroke detection in CT images.

This study investigates the detection of deepfake images and videos on social media platforms such as Instagramfor forensic analysis using hybrid-learning approaches. It highlights the critical importance of safeguarding privacy andauthenticity in digital media. The background draws attention to the growing threat posed by deepfakes, which posesignificant challenges across multiple domains, such as politics and entertainment. Existing methods often depend on visualfeatures specific to a dataset and struggle to generalize across different manipulation techniques. Moreover, mostapproaches focus exclusively on either temporal or spatial features, which limits their capacity to identify complex anomaliesinvolving fused facial features like the mouth, nose and eyes. Important solutions to these challenges include ConvolutionalNeural Network (CNN), Recurrent Neural Networks (RNN) and hybrid architectures that simultaneously capture spatial andtemporal information in deepfake content, such as Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM),Gated Recurrent Unit (GRU) and Vision Transformers (ViT). Additionally, this paper introduces a novel combination of artifactinspection and facial landmark recognition to enhance detection accuracy and employs Gated Recurrent Units (GRUs) andVision Transformers (ViT) for data augmentation thereby improving model robustness. The effectiveness of the proposedapproach is validated through experiments demonstrating substantially improved deduction accuracy, with improvementexceeding 1.5% across multiple datasets. However, several challenges remain, including limited robustness to noise, difficultyin detecting deepfakes in compressed video formats, and dataset imbalances issues. The proposed enhanced hybrid modelexhibits superior detection performance while maintaining adaptability across multiple datasets. Future research will focusstrengthening model generalization to effectively counter emerging deepfake generation techniques.

Abstract Structured light measurement is widely used in weld seam recognition due to its high precision and robust performance. However, in the automated guided welding process of narrow weld seam, robust detection cannot be achieved using traditional laser vision sensors due to complex environmental interference and the lack of distinct geometric contour information in the weld. To accurately extract narrow weld seam information and enhance image detection stability, we propose a semantic segmentation network for narrow weld seam based on spatio-temporal feature fusion. First, we utilize an enhanced laser vision sensor equipped with an auxiliary ambient light source to autonomously capture high-quality sequence images of narrow weld seam with laser stripe at the engineering site. Second, we design the spatio-temporal feature fusion narrow weld seam network (STFNet), and a topology encoder is introduced to extract the target's topological information. Subsequently, a spatial perception module and a temporal feature extraction module are proposed to capture the target's spatio-temporal information. The feature pyramid fusion module employs multi-level fusion to ultimately output precise weld detection results. Performance on our self-constructed NWSDataset demonstrates that our network effectively addresses real-time detection of narrow weld seam. It has been successfully deployed in submerged arc welding engineering applications, meeting practical welding requirements.

Pneumonia remains the most serious health menace in the world, particularly to children below the age of five. Early and correct diagnosis is important in minimizing morbidity and mortality levels. Chest X-ray (CXR) has been considered as one of the key diagnostic tools that offer invaluable information about the pulmonary abnormalities, like infiltrates and opacities. Nevertheless, manual review of CXR scans tends to be affected by inter-observer conditions and diagnostic lags, and inconsistencies due to environmental and staffing conditions. Recent advances in the domain of deep learning and Convolutional Neural Networks (CNNs) have offered a good fit to the problem of automation of pneumonia detection in CXR images. In this paper, the study of the use of ConvNeXt is described in detail and is compared and contrasted with classical CNN models, including AlexNet, VGG16, and ResNet50. Transfer learning was used to fine- tune five variants of ConvNeXt in order to classify pediatric CXRs as pneumonia or normal images. The ConvNeXt-Large model reached an unprecedented accuracy of 98.66 and exceeded its smaller counterparts and all the classical CNN models. The findings prove that the current CNN frameworks with transformer inspired design concepts can substantially increase the attribute extraction properties and the generalization perfor- mance. The fact that ConvNeXt has the potential to reduce the instances of misclassification is further supported by confusion matrix analysis. The results highlight the significance of transfer learning and larger and modern architecture in medical image classification. ConvNeXt-based models demonstrate good promise as effective and dependable clinical decision-support systems, and they can be used to help radiologists and help optimize diagnostic processes—especially where resources are limited. The paper ends with the research directions for the future, consisting of hybrid architecture, multimodal learning, and explainable AI to enhance trust and interpretability in the clinical field.

Accurate preoperative assessment of lesion size is crucial for selecting the appropriate endoscopic resection technique. However, the current assessment of lesion size still mainly relies on visual estimation, lacking objective measurement methods. To develop and validate an Endoscopic Virtual Ruler (EVR) based on image detection technology for objective measurement of lesion size before endoscopic treatment. Using computer image recognition technology and laser spot imaging principle, EVR was formed to detect the size of lesions. In vitro animal and human experiments were carried out to verify the accuracy and safety of EVR by comparing its measurement results with the actual size and the visual inspection results of endoscopists. In 30 in vitro tests, the measurement error of EVR was 0.08 ± 0.17cm (95% CI 0.01-0.14), and the relative accuracy of the measurement was 92.80% ± 5.50%( 95% CI 90.75-94.85%). In 58 clinical lesions, the mean error for visual estimation was 0.16 ± 0.66cm (95% CI-0.01 to 0.33), while EVR showed 0.12 ± 0.32cm (95% CI 0.04-0.21). EVR was significantly more accurate (85.68% ± 15.25%) than visual estimation (67.08% ± 22.59%, p < 0.001). EVR was more effective [48 (82.8%) vs 31 (46.6%), p = 0.001]. In the multivariable model, EVR-assisted measurement was independently associated with achieving clinically acceptable accuracy (OR 4.38, 95% CI 1.84-10.43, p = 0.001). EVR also demonstrated higher consistency in lesion size classification (Kappa = 0.764 vs. 0.522, p < 0.001). For lesions < 1cm, EVR misclassified only 12.5% as 1-2cm, significantly less than the 50% misclassification rate with visual estimation (p = 0.034). There was no laser damage side effect. EVR offers an accurate, safe, and objective measurement tool, which is helpful for the formulation of appropriate treatment decisions. ChiCTR2400085998.

Fish species recognition is essential for ecological studies, fishery management, and marine biology. Accurate detection and categorization are critical for preserving biodiversity, allowing scientists to track species distribution, identify invasive species, and analyze the effects of environmental changes. The fish sector is essential to any country's food and agriculture. Identification of species by the morphology process is both inaccurate and costly. However, the manual process of measuring important details like species identification, length, and quantity is difficult to capture, which shows the need for automation. The merging of automated systems and artificial intelligence has revolutionized this industry. Recent advancements in image detection systems based on machine learning and deep learning have been explored across various domains. Yet, applying state-of-the-art deep model Convolutional Neural Networks (CNNs) to identify the fish species’ complexity of season and location, and limited public datasets pose a challenge for the detection. Machine learning and deep learning use artificial neural networks to simulate how humans think and learn, efficiently automating similar monitoring applications such as species identification on land and in water. You Only Look Once (YOLO) is a state-of-the-art method for object detection based on deep learning. The goal of this study is to develop a deep learning system for recognizing fish species using the YOLO paradigm. The Fish-Pak dataset, which includes information on tropical fish farming in Pakistan, consists of 915 images against 6 targeted classes, freely available at the Mendeley data source. To ensure the suggested YOLO architecture's improved performance on the Fish-Pak data collection, we will conduct an experimental comparison with other versions of YOLO v3 and V4. The total accuracy of fish species identification using the proposed methods is 99%, with an mAP of 99.65%, top performance results as compared to existing literature.