Deep neural networks have achieved remarkable success in medical image classification, yet their deployment on resource-constrained clinical devices remains challenging due to computational and memory requirements. We present a novel thermodynamic pruning framework inspired by statistical mechanics that treats neural network weights as particles in a thermal system. By monitoring the thermodynamic temperature of weight configurations during training, we identify and remove redundant parameters while preserving critical representational capacity. Applied to ocular disease classification on the Ocular Disease Intelligent Recognition (ODIR) dataset, our method achieves 76.50% accuracy with 92.85% weight retention, representing an absolute improvement of +8.75% over the baseline (67.75%). Contrary to traditional pruning approaches that sacrifice accuracy for compression, our thermodynamic approach improves generalization while modestly reducing model complexity. The proposed method bridges the gap between theoretical physics and practical neural network optimization, offering a principled alternative to conventional magnitude-based pruning techniques.

A severe complication of diabetes that leads to permanently vision loss and early blindness give rise to the evolution of an ocular disease called as diabetic retinopathy. The major root cause of diabetic retinopathy is prolonged hyperglycemia that generally weakens the tiny retinal capillaries. The poor tiny retinal capillaries give rise to the development of microaneurysms and can completely block the flow of blood in the retinal vessels. This damaged retinal blood vessels induce the eruption of vascular endothelial growth factor VEGF which leads to the development of new fragile retinal blood vessels known as neovascularization. For this ocular diagnosing, fundus images play a crucial role. This paper represents a deep learning approach used for classification of diabetic retinopathy and normal fundus images from ODIR dataset integrating them with homomorphic filtering. Two convolutional neural network architecture, MobileNetV2 and VGG16 are hyper-tuned for image classification and disease detection. The final outcomes revealed better feature extraction and improved classification performance as the application of homomorphic filtering leads to enhanced contrast and illumination uniformly over the fundus images.

Recently convolutional neural networks (CNN) are applied to dermoscopic images to diagnose skin cancer. The dermoscopic images often have low contrast, making it challenging to differentiate between tissues or features. In some imaging modalities, scattered radiation can degrade image quality. Therefore, improvement in the images are required for more accurate diagnoses, as CNNs can rely on the enhanced images to identify the specific abnormalities. This study represents a hybrid approach for improving the disease detection in skin lesions by combining convolutional neural network with weighted guided image filter. The suggested algorithm retains the high quality features for classification. Xception network is utilized to extract features and to classify. The superiority of the suggested algorithm is validated on the HAM-10000 dataset. The proposed approach provides better accuracy compared to the original Xception model, thus assists early detection of seven categories of skin disease.

Skin cancer is a widespread health concern, affecting millions of people globally each year. Promptly and reliable examination of basal cell skin cancer, especially melanoma, can make a significant difference in patient outcomes. Melanoma, while less common than other skin cancers, is accountable for the maximum of deaths due to its aggressive nature. Traditionally, diagnosis relies on visual examination and biopsy, which can be subjective and sometimes delay treatment. This study explores an image- based method for assisting in the detection and classification of skin lesions. By analyzing patterns and features present in dermoscopic photographs, the approach aims to distinguish between benign and malignant lesions, with a focus on improving accuracy for challenging cases like melanoma. The results demonstrate that careful examination of visual characteristics in skin lesion images can support timely and more objective decision-making in clinical practice, potentially easing the diagnostic process for healthcare professionals and benefiting patients through earlier intervention.

Image fusion aims to combine redundant and complementary data from several source images. Creating strong features and discriminating models is the most difficult part of the process, as it improves the saliency information in the combined image. While several techniques are available to improve image quality, the optimal trade-off between computer vision processing and human observation remains to be found. This research proposed an image enhancement technique known as fusion-based shearlet counterlet enabled to produce high contrast in bright areas and improve visibility in dark areas of images, which combines the advantages of both NSST and NSCT. The Non-Subsampled Shearlet Contourlet Transform (NSSCT) serves as the backbone of the enhancement process, offering a multi-resolution and multi-directional analysis that captures intricate image details. The performance of the suggested method is evaluated using the performance metrics such as Structural Similarity Index Measure (SSIM) of 0.91, Peak Signal Noise ratio (PSNR) of 53.76 dB, and Mean Square Error (MSE) of 0.61. The model is evaluated on a normal brain dataset which shows superior performance over the traditional methods.

As the world continues to heal from the COVID-19 epidemic, the monkeypox virus has emerged as a new concern. New instances of monkeypox are reported daily, even though it is not as infectious or fatal as COVID-19. A worldwide pandemic is probable in the absence of adequate precautions. Medical imaging using deep learning (DL) approaches to diagnose medical conditions is currently showing encouraging results. The skin of humans afflicted with the monkeypox virus can facilitate early diagnosis of the disease since image processing techniques have been employed to enhance understanding of the condition. This work aims to identify monkeypox ailments during a potential pandemic by analyzing skin lesions using deep-learning techniques efficiently and securely. Improving the transfer learning model with hyperparameters allows for the development of a hybrid deep learning model within the convolutional neural networks (CNN) framework. For this study, we used multiple types of CNN models, including ResNet-18, ResNet-50, Xception, SqueezeNet, and EfficientNet-b0. In our study, the monkeypox skin image database (MSID) is used and enhanced using Brightness preserving dynamic histogram equalization (BPDHE) for the improvement of contrast of images. The proposed algorithms are assessed according to the performance metrics. Experimental findings showed that ResNet-50 is the most effectual CNN technique in the recognition of monkeypox.

Object detection is considered as detection of an object that appears in an image or video. This system helps in identifying and classifying the objects present, such as whether the object is a car, person, dog, etc. A human brain can efficiently perform this task and does not require much time and processing, but it is not the same case for a machine. A machine requires much time and data processing to identify what object is present. In order to detect an object inside the image, there are various ways to achieve using computer vision techniques. Although there have been a large number of studies that thoroughly explored various types of methods for object detection. Here, this article compares four mainstream object detection models to detect objects inside an image tracing the evolution from traditional machine learning-based methods to contemporary deep learningdriven approaches. The analysis is drawn based on different published literature for machine learning and deep learning based techniques. This review attempts to assist academics and practitioners in choosing and developing object detection methods for various application domains by providing critical analysis and comparative insights.

Conventional image processing techniques identify the features of the diseases manually, which demonstrate poor efficacy and poor detection rate. To solve this issue, this work has suggested an improved convolutional neural network (CNN) architecture that is utilized for multi-class recognition of plant disease images. Initially, the traditional network, such as GoogLeNet and EfficientNet, is utilized for feature extraction and categorization. Secondly, the features of the traditional network are enhanced by hybridizing with low low-light image enhancement algorithm. The experiment utilizes the sugarcane leaf disease dataset that is available on Kaggle. The results outperform the original CNN models. At the same time, the recommended techniques have converged with a faster speed than traditional approaches.

Alzheimer's disease is a prevalent nervous system illness among the elderly population, and its likelihood of developing rises with age. One potential non-invasive measurement method that is frequently used to gauge changes in brain impulses is the electroencephalogram. To differentiate Alzheimer's patients, EEG waves are analyzed. Several studies assessed the clinical significance of Alzheimer's disease. In particular, recently developed evaluation techniques based on machine learning bring the fact to light by extracting features from EEG signals. Optimizing the network parameters of machine learning models using optimization algorithms can also lead to enhanced performance. Consequently, this study effort presents an optimized machine learning model for deep belief network-based Alzheimer's disease identification.. Particle Swarm Optimization is included in the proposed work to optimize the deep belief network parameters. The results are compared with the traditional deep belief network and illustrate that the suggested method outperforms the traditional approach.

The analysis of retinal images plays an instrumental role in diagnosing Diabetic Retinopathy. This progressive disease can cause blindness which can be inhibited with earlier detection. A robust approach is presented in this paper for blood vessel segmentation by integrating a matched filter with Kirsch’s template with hysteresis thresholding. The proposed approach involved three steps: preprocessing, blood vessel extraction, and post-processing for vessel extraction. This approach achieved more accurate results than the original Kirsch’s template for specificity, sensitivity, and accuracy on the DRIVE dataset.

Traffic congestion is prevalent in many major and medium-sized cities throughout different countries in contemporary society. In traffic images, various multi-sized vehicles are tightly clustered together and obstructed from one another. Identifying vehicles in such instances is crucial for urban traffic surveillance, safety monitoring, and legal concerns but it also presents major challenges. The remarkable detection accuracy and efficiency of deep learning-based systems have led to their recent and extensive use in vehicle identification. There are significant advanced YOLO models with different backbone architectures and frameworks developed for vehicle detection. Yet, the performance of YOLO variants are facing the challenges of handling false detection against occluded and densely sophisticated scenarios. The proposed model is developed to address such types of limitations, for example; dynamic illumination, noisy images, and scale sensitivity to improve the vehicle detection rate in different traffic scenarios and varying weather conditions. The proposed study employs an improved YOLOv4 to identify moving vehicles in different lighting conditions including daylight, cloudy, rainy, and night. For hybridization, three techniques are utilized such as the Multiscale Retinex, Dual tree complex wavelet transform (DTCWT), and Pulse Coupled Neural Networks (PCNN). The DTCWT is employed for multiscale decomposition and to denoise the complex high frequency subband information, then the denoised subbands are reconstructed into a denoised image. The Multiscale retinex is utilized to reduce the halo artifacts on high-contrast edges and maintain the balance with dynamic range compression and color reproduction. The synchronizing pulse burst property of PCNN is used to detect the isolated noisy pixels and modify the detected noisy pixels. From the results it is worth noting that the developed model surpasses state-of-the-art methods in sunny, night, cloudy, and rainy modes. The proposed method using the DTCWT technique can detect the vehicles with mAP of 91.09% and 35FPS.

Blockchain is a decentralized architecture with built-in security to improve the trust and integrity of transactions. Blockchain can facilitate authentication and authorization without utilizing any trusted authority. There is no single authority that governs how the rules will be applied. Anyone is free to join the public blockchain network. Bitcoin is an example of public blockchain. Consortium blockchain is considered as partly decentralized architecture that can be open or private. Hyperledger and R3CEV are the examples of consortium blockchain. In private blockchain, nodes are restricted and have strict authority management on data access. Keeping in view of different blockchain technologies, this chapter has proposed four different applications of blockchain technology. The first one is detecting product genuineness using blockchain. The second is an application of blockchain in citizen participation. The third one is implementation of security in cloud computing using blockchain. The fourth is e-voting and event registration using blockchain technology.

Brain tumors can be difficult to diagnose, as they may have similar radiographic characteristics, and a thorough examination may take a considerable amount of time. To address these challenges, we propose an intelligent system for the automatic extraction and identification of brain tumors from 2D CE MRI images. Our approach comprises two stages. In the first stage, we use an encoder-decoder based U-net with residual network as the backbone to detect different types of brain tumors, including glioma, meningioma, and pituitary tumors. Our method achieved an accuracy of 99.60%, a sensitivity of 90.20%, a specificity of 99.80%, a dice similarity coefficient of 90.11%, and a precision of 90.50% for tumor extraction. In the second stage, we employ a YOLO2 (you only look once) based transfer learning approach to classify the extracted tumors, achieving a classification accuracy of 97%. Our proposed approach outperforms state-of-the-art methods found in the literature. The results demonstrate the potential of our method to aid in the diagnosis and treatment of brain tumors.

A brain tumor is one of the deadliest neurological diseases developed in the human brain. Gliomas are the most common type of brain tumor which are originated from the glial cells of the brain and are treated as the most treacherous tumors due to aggressiveness and heterogeneous structure. Early detection of gliomas increases the survival rate of the patients. In this paper, a hybrid deep neural network is proposed that uses inception v2 network hybridized with 16 new layered segmentation nets. The network is tested on BraTs 2020 and BraTs 2017 multi-parametric MRI (mPMRI) dataset to detect the whole tumor, and for the detection of tumor core (TC) and the edema, fast fuzzy C-means (FFCM) method is used. The proposed method attains an accuracy of 99.45% and a Dice similarity coefficient (DSC) of 89.74% for the whole tumor, an accuracy of 99.36%, a DSC of 86.47% for the tumor core and accuracy of 99.45%, and a DSC of 85.22% for the edema.

According to a scientific study, eyes are the best predictors of numerous disorders including glaucoma, diabetic retinopathy, hypertension, and stroke. An ophthalmologist can learn about the problems by looking at the segmented retinal blood vessel network. The goal of this study is to offer ophthalmologists with reliable segmented retinal blood vessels to help them pinpoint the issue. This work put forwards an automated method of vessel extraction by incorporating curvelet-based enhancement with the Coye algorithm. Further, the segmentation performance is fine-tuned by embodying a pair of complementary gamma functions (PCGF) for contrast improvement. The suggested approach is evaluated on DRIVE and STARE databases and shows outstanding results as compared to state-of-the-art algorithms.

Gait recognition has become one of the furthest promising behavioral biometric techniques for identifying individuals. Gait recognition models are capable of identifying humans at a distance based on their walking manner without their permission or interference. However, it is usually noted that the performance of a gait recognition approach will drop drastically in the presence of covariates such as carrying conditions, clothing conditions, and variations in the view angle. Therefore, it is necessary to develop a robust gait recognition system in order to identify the most significant gait features. In this paper, we introduced a recently developed hybrid whale and gray wolf optimization algorithm (WGWOA) for determining the optimal subset of gait features by combining the advantages of whale optimization and gray wolf optimization techniques. Moreover, we employed principal component analysis (PCA) to extract the essential gait features from the gradient gait energy image (GGEI) and random forest (RF) approach to classify the optimal gait features. The proposed method has been assessed on the publicly available largest multi-view CASIA-B and OU-MVLP benchmark datasets. Experimental results indicate that the proposed model achieved an accuracy of 99.25%, 98.39%, and 97.97% under normal, carrying a bag and wearing coat walking conditions, respectively on the CASIA-B dataset. Furthermore, the proposed model achieved an accuracy of 97.63% under normal conditions on the OU-MVLP dataset. The comparative results also confirm that the proposed algorithm is superior to the contemporary approaches.