Embroidery cultural relics often suffer from missing stitches and colour fading due to environmental and human factors. Traditional manual restoration is time-consuming (over 200 hours per piece) and prone to secondary damage. Existing deep learning methods struggle with structural distortion (>15%) and poor semantic consistency in complex textures like gold/silver threads and three-dimensional embroidery. To address these issues, we propose an advanced embroidery image restoration method. A DenseNetwork with channel and spatial attention achieves 94.25% accuracy, 93.47% recall, and 96.59% specificity for classification. Restoration uses an improved GAN with dilated convolutions, attention modules, a mask-guided discriminator, and joint loss. On datasets with 20-30% masking, the SSIM reached 0.971 (vs. 0.873 for traditional GANs). At 40-50% masking, the FID dropped to 17.33 (vs. 20.14). The model is efficient, requiring only 10.52G FLOPs, 5.58M parameters, and 0.62s per image. This method enables high-quality, efficient restoration of embroidery artefacts.

Soft set theory is a useful mathematical paradigm for dealing with imprecise information and controlling uncertainty. By offering parameterised family of subsets defined over a universal set of parameters, it allows for flexibility in handling uncertainty in a variety of situations. Using soft set theory to express ambiguous problem parameters is still difficult, despite its benefits. With an emphasis on parameter reduction to expedite the selection of ideal objects, this paper explores the function of soft set theory in decision making. Parameter reduction increases efficiency by streamlining decision-making procedures while preserving crucial information. Additionally, a comparison table built using fuzzy soft set is used in a novel approach to object detection from imprecise data. In unpredictable situations, this approach increases decision-making accuracy and dependability. The study illustrates the potential of soft and fuzzy soft sets to improve decision-making processes through real-world situations, showcasing their adaptability and resilience in dealing with uncertainty.

Diabetic retinopathy (DR) is a leading cause of blindness in diabetic patients, emphasising the need for early detection and efficient classification techniques. Existing methodologies focus on conventional deep learning models that often struggle with incremental learning and may overlook crucial exudate-based features. These models face challenges in accurately segmenting complex features in fundus images, leading to sub-optimal classification results. We propose an innovative exudate statistics-based incremental deep learning model (ESIDLPD) to overcome these limitations. Utilising fundus images, the model accurately extracts a wide range of exudates and segments complicated features using the UNet architecture. Segmented blocks are transformed into multi-domain features and are classified into DR classes through an LSTM-based RNN process, supplemented with a VARMAx process for enhanced prediction. This approach refines categorisation, improving precision by 4.5%, accuracy by 3.5%, recall by 1.9%, AUC by 2.5%, and specificity by 1.5%, while reducing diagnostic delay by 10.4%.

Inderscience is a global company, a dynamic leading independent journal publisher disseminates the latest research across the broad fields of science, engineering and technology; management, public and business administration; environment, ecological economics and sustainable development; computing, ICT and internet/web services, and related areas.

Image recognition has become essential in biomedical research, particularly for disease diagnosis and biomarker detection. While convolutional neural networks (CNNs) have achieved success in image tasks, their performance declines with complex, noisy biomedical data. This study compares ResNet and traditional CNNs in antibody immune detection. ResNet's residual connections effectively address vanishing gradients in deep networks, improving training stability and maintaining consistent performance. Its superior feature extraction captures subtle image variations, achieving high classification accuracy under noisy conditions. ResNet also shows strong robustness across different network depths. Experimental results demonstrate that ResNet outperforms conventional CNNs in detection accuracy, especially with specialised biomedical images, and holds significant promise for clinical application. The findings indicate that ResNet is a more reliable and accurate framework for biomedical image recognition, especially in complex and high-noise environments, offering advantages in both research and practical diagnostic contexts.

Wireless sensor networks (WSNs) are essential in environmental monitoring, surveillance systems, and smart cities. Achieving optimal coverage in WSNs is a fundamental challenge for effective event detection and monitoring. This research introduces an adaptive crossover mutation differential evolution (ADE) approach to address the node coverage planning problem in WSNs. The ADE method optimises sensor placement to maximise coverage while minimising the number of required sensors. It incorporates enhanced adaptive strategies for crossover, mutation, and re-initialisation to improve coverage efficiency. The ADE approach is evaluated using test suite functions and compared against existing optimisation strategies using various metrics for validation. Experimental results demonstrate the superior performance of the proposed approach in maximising node coverage in WSNs. The ADE approach presents a novel and effective solution for optimising coverage in WSNs with potential practical applications.

Protecting our servers and machines is vital since they store our data and resources. Attackers use the latest tools and technologies to launch the attack. Among, several cyberattacks, DDoS attacks are the worst. DDoS attackers employ a variety of methods to exploit machines and consume all server resources to block authorised users. Current DDoS detection methods depend on network topology, cannot detect all types of attacks, use outdated or invalid datasets, and require powerful and expensive infrastructure hardware. In our research, we filter non-legitimate traffic and use machine learning classifiers to predict attack types from attack traffic. These two processes together reduce attack volume and identify attack type to provide a comprehensive DDoS protection and enable targeted reaction and mitigation. We used CIC DDoS 2019 data for the experiment. It records MSSQL, PortMap, LDAP, NetBIOS, Syn, UDP, UDPLag, and benign traffic attacks. Experiments yield promising and satisfying results.

Glaucoma is the second leading cause of blindness and the first leading cause of irreversible blindness. The main types of the disease are open-angle and angle-closure glaucoma. In people with angle-closure, the anterior chamber angle is narrow, which leads to a rising intraocular pressure, and consequently, optic nerve damage, causing vision loss. Since it is irreversible, an early diagnosis is essential. So, angle classification is fundamental for diagnosis. Anterior segment optical coherence tomography is one of the imaging tests used to diagnose the disease. In addition to the fact that no eye contact is necessary, this test provides a fast way to analyse the anterior chamber angle, to classify it as open or closed. We propose the anterior chamber angle classification method based on visual feature extraction, using deep neural networks in this work. In a multilevel architecture, different pre-trained CNNs are adjusted to extract deep features and train two classifiers. The best model extracted visual features from the anterior camera angle in the experiments, and achieved a sensitivity value of 1.000 as the best result.

This manuscript studies the stability for a class of Cohen-Grossberg neural networks (CGNNs) with variable delays. By practicing the scheme of Lyapunov function (LF), M-matrix (MM) theory, homeomorphism theory and nonlinear measure (NM) method, a new sufficient condition is obtained to ensure the existence, uniqueness and global exponential stability (GES) of equilibrium point (EP) for the studied network. As the condition is independent to delay, it can be applied to networks with large delays. The result generalises and improves the earlier publications. Finally, an example is supplied to exhibit the power of the results and less conservativeness over some earlier publications.

Leukemia is a disease that has no defined etiology and affects the production of white blood cells in the bone marrow. Young cells or blasts are produced abnormally, replacing normal blood cells (white, red blood cells, and platelets). Consequently, the person suffers problems in transporting oxygen and infections combat. Acute leukemia is a particular type of leukemia that causes abnormal cell growth in a short period, requiring a quick start of treatment. Classifying the types of acute leukemia in blood slide images is a vital process, and a system of assisting doctors in selecting treatment becomes necessary. This paper presents an ensemble approach using four convolutional neural networks (CNNs) -Alert Net-RWD, Resnet50, InceptionV3, and Xception. These CNNs, individually, demonstrated that are efficient in differentiating between the two types of acute leukemia -Acute Lymphoid Leukemia (ALL) and Acute Myeloid Leukemia (AML) -and Healthy Blood Slides (HBS). We verified that the union of these four well-known CNNs improve the hit rates of current techniques from the literature. The experiments were carried out using 18 data sets with 3,293 images, and the proposed CNNs ensemble achieved an accuracy of 96.17%, and precision of 96.38%.