While the use of artificial intelligence in ophthalmology has increased over the past few years, inflammatory diseases like choroiditis remain widely unexplored. A delayed diagnosis in such cases can cause severe complications that can lead to visual disability, and hence, automated diagnosis systems can be of great importance. In this study, we present a systematic, evidence-driven methodology for designing a lightweight choroiditis classifier using a real-world small dataset with a constraint of low resource availability. This hybrid framework consists of two components. The first feature extraction component extracts features using the MobileNet-V3-Small model. The second component, namely the classification component, utilises the Support Vector Machine as the binary classifier. This optimal combination was identified through systematic comparative experiments. Statistical testing confirms the robustness of the classifier selection. The model gives a cross-validation accuracy of 94% and a held-out test accuracy of 97.06% with a training time of approximately 3 minutes for the end-to-end pipeline on a carefully collected and previously introduced choroiditis dataset. Being lightweight and computationally efficient, this model is a suitable candidate for future development into a reliable computer-aided diagnosis tool that could assist experts in reviewing images, providing telemedicine care, and prioritising patient appointments based on the initial results of the automated systems. 

The rapid development in deep learning-based generative softwares and image rendering tools has led to generation of massive photorealistic digital content – Fake images, fake videos, fake speech, etc. Such fake digital media may result in communication of misinformation, forgery of digital data, and losing trustworthiness in the information source. This poses a significant challenge to the field of digital forensics’ techniques, to which our present work attempts to make a contribution, by addressing the problem of differentiating AI-generated images from real photographs, using transfer learning and multi-branch fusion model. We propose a multi-branch model that integrates two pre-trained Vision Transformer models (DINO (self-distillation with no labels) and Contrastive Language – Image Pretraining (CLIP)) to extract complementary global features, along with a forensic and a hand-crafted feature branch, which extract low-level discriminating cues. These features are complimentary to each other and hence contribute in improving the robustness and performance of the model. These features from the four branches are adaptively weighted and combined by a cross-attention module, to give a fused and rich embedding. The model is further optimized by using augmentation-invariant loss, center loss and supervised contrastive loss in addition to the cross-entropy loss function. This framework achieves improved accuracy of 95.83% on PRCG dataset, 96.79% on CIFAKE dataset and 99.15% on GenImage dataset as compared to baselines. It also achieved stable cross-generator performance and enhanced robustness against real world corruptions like Blur, Noise, Compression, and others. The experimental results show a good separability between the classes, and enhanced performance on publicly available datasets.

Currently, image compression algorithms are an integral part of modern information systems in various industries and spheres of human activity, including telecommunications, medicine, artificial intelligence, and defense technologies. This paper deals with a novel image compression method based on nonlinear multiscale decomposition with fractional-rational approximation, providing a compact representation of image components while preserving reconstruction quality. The proposed algorithm consists of five steps, including image preprocessing, discretization, nonlinear multiscale decomposition, quantization, and arithmetic compression. The method was evaluated using 1000 test images and demonstrated an average compression ratio of 13.2, with reconstructed image quality of 41.5 dB, outperforming classical wavelet-based approaches under comparable  conditions  (approximately by 8-10% in average). The computational complexity of the proposed algorithm remains suitable for practical implementation, making it a promising solution for efficient image compression in modern digital systems.

Microplastics are tiny particles made of plastic that are a significant source of pollution in the sea, and are dangerous to both the environment and human health. Nevertheless, the existing detection techniques are not robust and general in dynamic underwater settings with varying microplastic shapes and sizes. To overcome these issues, a high-resolution underwater microplastic segmentation framework was implemented that integrates HCTSpeckle and Vision Transformer Encoder–Sandwich Decoder Network. Initially, underwater sensors record continuous visual images. The captured images are pre-processed with HCTSpeckle, a CNN-Transformer denoising network, for removing speckle noise, and it retains important structural information through the use of hybrid convolution-transformer blocks and double residual interactions. The denoised images were contrasted with the Lightweight YUV Transformer-based Network, in which multistage squeeze-and-excitation fusion is used to improve the visibility of object boundaries. The enhanced images are subjected to a hybrid Vision Transformer- Sandwich Decoder to produce precise underwater microplastic segmentation, where the ViT encoder captures high-level features that are globally correlated without distorting positioning information in space by patch embeddings and self-attention mechanisms. They are decoded using a Sandwich Decoder Network, which learns both local and global dependencies. Also, ranking and region-based pooling priorities fine edges and microplastic structures, whereas the pixel-wise segmentation head precisely categorizes the identified microplastics into fiber, film, pellet, and fragment types. The proposed approach attains the pixel accuracy of 0.97 and specificity of 0.98 with a Dice Coefficient of 0.934, which contains the effective segmentation result of underwater microplastic images. These findings ensure the framework efficiently identifies and categorizes various types of microplastics in diverse underwater sceneries.

Skin cancer is one of the deadliest kinds of cancer because of it’s due to its visible onset and the potential for rapid progression and spread. Oncologists employ several techniques, including imaging and biopsies, to determine whether skin cancer is present, but these are labor-intensive and time-consuming. Developing an automated and accurate framework is essential for early skin cancer detection, which greatly increases the probability of a successful treatment and recovery. Therefore, to address the aforementioned problems, this study aim to propose a novel Deep Learning (DL)-based multi-class skin cancer classification (MC-SC2) model for automatically diagnosing skin cancers using dermoscopic images. Initially, the dermoscopic images are preprocessed using a Gaussian Filter (GF) to effectively remove noise. To address the class imbalance issue, we applied advanced augmentation techniques to oversample minority classes, ensuring a uniform class distribution and enhancing the model's ability to generalize. To extract the complex and dominant features, we propose a novel DL-based model called BAM-EfficientNet (Bottleneck Attention Module with EfficientNetB7). In BAM-EfficientNet, we replace each Squeeze-and-Excitation (SE) attention module with a BAM in the traditional EfficientNetB7; this modification enables the network to concentrate on the most relevant regions in the images. The extracted features fed into the proposed Modified Stacked Capsule Network (MSCNet) to classify a skin lesion image as AKIEC, BCC, BKL, DF, MEL, NV, and VASC. In the proposed MSCNet employs the Disperse Dynamic Routing (DDR) algorithm to improve capsule networks performance, and network's initial weights and biases are fine-tuned using Pelican Optimization Algorithm (POA) to further enhance performance. The HAM10000 dataset is used to assess the suggested model. The findings indicate that our approach beats current methods with an accuracy of 99.21% in classifying the seven different types of skin cancer, yielding substantial outcomes. These results demonstrate the potential of the suggested model as a quick, precise, and useful tool for early skin cancer diagnosis, providing important assistance in diagnosing skin cancer for medical professionals.

The rapid growth of motor vehicles in urban areas has led to an increasing demand for parking facilities and requires tariff policies that are more adaptive to real field conditions. This study aims to develop a parking revenue potential classification model to support the formulation of progressive parking tariff policies using a data-driven and spatial analysis approach. The dataset includes vehicle attributes, parking volume, and temporal parking usage patterns. Parking revenue potential is categorized into low, medium, and high classes using a quantile-based approach. Unlike most previous studies that focus on parking occupancy prediction, this study proposes a revenue-oriented spatiotemporal classification model integrating spatial coordinates and temporal parking patterns to support adaptive tariff policy formulation. The Random Forest algorithm is applied to classify parking revenue potential into low, medium, and high categories, achieving an accuracy of 90.24% for two-wheeled vehicles and 89.02% for four-wheeled vehicles. The classification results are integrated into an interactive Streamlit-based dashboard that visualizes the spatial distribution of parking revenue potential and enables simulations of progressive tariff adjustments based on spatial zones and temporal conditions. The proposed system functions as a decision support tool for parking management, aiming to improve operational efficiency and sustainably increase regional parking retribution revenue.

Early polyp detection is vital in avoiding colorectal cancer, a top contributor to cancer-related deaths globally. Automated polyp detection has been greatly improved by deep learning, but extensive deployment tends to be hampered by patient data privacy fears. For this purpose, we suggest a federated learning (FL) paradigm that enables decentralized model training without sharing raw patient data while having high diagnostic accuracy. We first performed an extensive evaluation using DL models on a polyp dataset that we gathered. MobileNetV2 was the best performing model as per important metrics like accuracy, precision, recall, and F1 score. In order to further expand its representational power, we incorporated a transformer module into MobileNetV2 so that the model can better capture long-range dependencies and context information. Our new Transformer-Enhanced MobileNetV2 model was then implemented on several simulated clients in a federated learning scenario. This configuration enabled training over decentralized clinical data without violating patient privacy. We utilized standard FL algorithms for model averaging and evaluated the system in accuracy, precision, recall, F1 score, and convergence time. The devised approach performed extremely well with an accuracy of 98.37% and an F1 score of 0.971 while converging effectively in 27-33 rounds. These findings imply that a promising path for safe and effective medical image analysis is to combine transformer designs with lightweight models in a federated condition.

Nephrolithiasis (kidney stone disease) is a common urological disease that has a high clinical and economic impact. The early diagnosis is needed to avoid complications like obstruction of the ureter, infection, impaired kidney functioning. Traditional imaging modalities, such as ultrasonography, kidney-ureter-bladder radiography, and non-contrast computed tomography, are common but have a number of limitations, specifically their operator dependence, radiation, and low sensitivity to small or radiolucent stones. This review follows a PRISMA-based methodology to conduct a systematic review of studies published between 2015 and 2025 on the topic of computational intelligence methods such as artificial intelligence, machine learning, and deep learning to detect kidney stones based on medical images. Major scientific databases were considered in studies according to imaging modality, preprocessing method, model architecture and performance measures. Deep learning models, especially, Convolutional Neural Networks and U-Net-based frameworks, are highly effective in detection and segmentation tasks and have been reported to have accuracy of 86 to 99.9 percent, Dice coefficients over 0.85 and AUC of up to 0.99 in controlled data. Hybridization to combine ML classifiers, including Support Vector Machines, further improves the performance of classification. Yet, these outcomes are commonly limited through small datasets, class imbalance, external validation, and overfitting, which have an impact on real-life generalization. The use of computational intelligence has greatly improved the detection of kidney stones by enhancing automation, precision, and reproducibility. However, there are still major issues, such as the standardization of the dataset, interpretability of the models, and limitations to the clinical implementation. Explainable AI, federated learning, and 3D volumetric analysis should be prioritized in future research to create diagnostic systems.

It is still hard to accurately find bone fractures and figure out how bad osteoarthritis (OA) is from X-ray pictures because of the complicated anatomical differences and the lack of contextual modelling in standard deep learning methods. This research presents XrayBoneNet, a hybrid deep learning system that combines Convolutional Neural Networks (CNNs), Transformer-based global feature modelling, and attention mechanisms for concurrent fracture identification, OA grading, and localization. The model has two heads: one for binary fracture classification and one for multi-class OA staging. It also has a bounding box regression head for accurate localization. To improve training efficiency and performance, a hybrid optimization technique that uses Bighorn Sheep Optimization (BSO) for global exploration and Logarithmic Mean Optimization (LMO) for fine-tuning is developed. The Bone Fracture Multi-Region X-ray dataset shows that XrayBoneNet works better than state-of-the-art models like ResNet50, DenseNet121, and Vision Transformer. It has 96.8% accuracy in fracture detection, 94.5% accuracy in OA classification, and an Intersection-over-Union (IoU) of 0.87 for localization. The suggested system offers an effective and comprehensible alternative for automated radiological diagnosis.

Road safety depends on both a driver’s emotional cognitive state and physical driving behavior, yet most existing systems rely on a single modality, limiting real-world reliability. This paper presents HECANet (Hybrid Evolutionary Cross-Attention Network), a multimodal framework that integrates EEG-based emotional cues and vehicle telematics behavior for robust driver risk assessment. EEG signals are modeled using a PSO-optimized CNN–LSTM to capture spatiotemporal emotional patterns, while telematics data are analyzed using a GA-optimized XGBoost model to identify safe, distracted, and aggressive driving behaviors. A Differential Evolution–optimized cross-attention fusion layer effectively aligns emotional and behavioral features, enabling interpretable emotion–behavior interaction modeling. The fused representation produces a driver safety score and risk probability, with K-Means clustering used to categorize drivers into Safe, Caution, and Risky groups. Experimental results achieve 94.7% accuracy and a 0.94 macro F1-score, demonstrating that joint emotion–behavior modeling significantly enhances driver risk prediction for intelligent transportation and fleet safety applications.

Image fusion is the method of combining the features of different images into one to get a more informative or high-quality image. Among its various types, multi-modal image fusion is a crucial one where images obtained using sensors receptive to different light radiation are integrated into one final image. Infrared (IR) and Visible (VIS) Image Fusion (IVIF) is one such popular fusion technology. In IVIF, visible sensor produces clean texture and structure information, while it is sensitive to illumination and occlusion. IR sensor, though vulnerable to noise, captures salient targets that emit thermal radiation. The contrasting properties of the two images can be exploited by producing a fused image that both highlights the prominent target as well manifests detailed information. First, the acquired IR and VIS source images are each decomposed using the Gaussian blur filter into base (low-frequency) and detail (high-frequency) components. As opposed to the conventional way of concatenating the respective base and detailed components of the source images, a new technique of combinative concatenation has been performed providing a comprehensive set of 6 unique features to perform fusion. The proposed combinative concatenation is mathematically formulated, illustrating how cross-modal feature generation improves the retention of information and enhances modal complementarity. Weighted Sum (WS), Principal Component Analysis (PCA) and Laplacian Pyramid (LP) have been used for the fusion process. The 6 unique features extracted are fused in 20 different ways considering all combinations to provide fused results with different properties. Finally, a set of 4 statistical analysis methods are applied to identify the best fusion strategy. As a highlight, this paper has assessed these fusion strategies over live images captured using a Near-Infrared (NIR) and VIS camera depicting different illumination conditions (bright, dim and dark), and its effects over the fusion performance are assessed in comparison to fusion of similar images from an existing dataset.

Skin diseases range from mild conditions to severe threats, such as melanoma highlighting the critical importance of early and accurate evaluation for effective treatment. Traditional diagnostic methods that depend heavily on visual inspection and biopsy are often prone to delays and susceptible to human error underscoring the need for more efficient and reliable approaches. This study introduced a computerized system for detecting skin diseases using deep learning techniques and classification, with a focus on enhancing diagnostic efficiency and accuracy. The approach begins with the comprehensive preprocessing of skin images including illumination adjustment, elimination of artifacts via morphology closing and edge improvement using an unsharp filter. These steps improve the image clarity and prepare the data for accurate analysis. To address variations in lesion size and boundary irregularities a Fuzzy k-means clustering technique segments the affected skin regions, ensuring adaptable detection across diverse skin conditions. The classifier Deep-MobileNet integrates the lightweight MobileNet architecture with Squeeze-and-Excitation (SE) blocks enabling it to effectively differentiate between healthy and diseased skin. SE blocks enhance the model’s ability to capture spatial dependencies thereby improving classification precision. To further optimize the classifier’s performance hyperparameters are tuned using the adaptive deer hunting optimization (ADHO) algorithm, which accelerates convergence and boosts model efficiency. By providing an automated efficient solution this approach has the potential to assist healthcare providers in diagnosing skin diseases quickly and reliably thereby supporting timely and effective treatment interventions. The efficiency of the suggested method is evaluated based on accuracy, sensitivity, specificity and F1-Score. The experimental outcome showed that recommended approach attained a maximum accuracy of 91.96%.