Fusion: Practice and Applications

Fusion techniques on enhancing the efficiency of Long Short-Term Memory (LSTM) networks are dominating across a variety of domains. To handle sequential data while integrating from various sources is often challenging using LSTM techniques. Fusion methods that integrate different models enhances LSTM’ ability to handle complex correlations in the data. This paper examines early, late and hybrid fusion techniques. The study provides fusion approaches to enhance LSTM networks to efficiently handle complex multimodal data across self-navigating models. The findings reveal that the hybrid fusion techniques outperform traditional methods in terms of accuracy and generalization of various tasks. This paper proposes the Gated Recurrent Fusion (GRF) approach to demonstrate its performance to handle multimodal and temporal models in a supervised recurrence. The findings report 10% enhancement in terms of precision rate

Adverse Drug Reaction (ADR) is a significant global public health issue and the main cause of death. Generally, the effects of ADR are complex. Clinically, they can cause major patient damage and, in some cases, death. Besides, this outcome in significant healthcare costs financially owing to enlarged hospital visits, extra treatments, and harm to productivity. Therefore, early recognition and mitigation of ADRs are vital for the patients. Enhancing the early detection of ADRs and deadliness could severely reduce the harm to patients, improve patient safety, decrease healthcare costs, and increase the efficacy of the drug development procedure. Conventional pre-clinical toxicity tests are expensive, time-consuming, and frequently fail to forecast human-specific toxic effects. Artificial Intelligence (AI)-based deep learning (DL) has been quickly adopted in numerous areas, with healthcare, for its latent to manage huge datasets, find out patterns, and generate predictions. This study presents a new Adverse Drug Reaction Detection through Deep Learning and Improved Red-Tailed Hawk Algorithm (ADRD-DLIRTHA). The main intention of the ADRD-DLIRTHA model is to enhance the detection and classification process of ADR using advanced hybrid and optimization techniques. At first, the data normalization stage applies z-score normalization for converting input data into a beneficial set-up. Furthermore, the proposed ADRD-DLIRTHA method designs a convolutional neural network and long short-term memory (CNN-LSTM) technique for the classification process. At last, the improved red-tailed hawk (IRTH) algorithm-based hyperparameter selection process has been applied to optimize the classification results of the CNN-LSTM system. A wide range of experimentation was led to authorize the performance of the ADRD-DLIRTHA system. The simulation results specified that the ADRD-DLIRTHA model emphasized advancement over other existing techniques

The intersection of the Electronic Health Records (EHR) is the main factor that makes healthcare delivery and the patient outcomes better. On one hand is the seamless combination of the EHR systems of different departments in preserving data security and privacy is a great achievement, but on the other hand, the integration of the EHR systems of different departments while maintaining data security and privacy is still an important concern This paper suggests a new blockchain-based secure framework that may be used to improve the interoperability of patient data among the EHR systems. The blockchain technology, which is immutable and decentralized, supports the major principles of the framework such as data integrity, security, and privacy.  The proposed model comes with a strong recommender system, which makes the patient-doctor consultations, specialist suggestions, and the laboratory test requests according to the symptoms and doctors' recommendations more efficient. Thus, the system, when linked with Google Maps, recognizes local laboratories, and allows for direct test requests; consequently, the healthcare process is made more effective. The analyzed system optimizes the data exchange, protection, and the functionality of the informational system in contrast to the current EHR systems. It is therefore apparent that this blockchain-based technique is one that can efficiently address the challenges of EHR integration and therefore goes down well with the future of secure and efficient healthcare systems. Assessment of the framework demonstrates the effectiveness of the proposed adjustments in various aspects, such as data security and data compatibility and system; tests affirm the improvement of the user’s satisfaction and the improvement of the data management

Presenting the network architecture EMOTE-Net is a method of enhancing the face emotion recognition and classification in video data for this work. The suggested model merges the use of DenseNet to extract features with the SVM (support vector machine) to categorize the data by specifying SVM here. This feature of EMOTE-Net is highly outstanding because SVM and DenseNet are combined and are thus capable of sophisticated classification and effective feature extraction. The first process to come in methodology is preprocessing of video data. Bounding Box detection is able to extract regions that are of interests (ROIs) and that Densenet is great at the feature representation with high dimensions. Henceforth, feed these features into a classifier from SVM for intelligent categorization. Evaluation has provided clear evidence regarding the efficiency of this model, which has obtained the accuracy of 0.9890, precision of 0.9900, sensitivity of 0.9877, specificity of 0.9972, and F1 score of 0.9886. The pertinence of EMOTE-Net to real life applications, such as video analytics, human-computer interaction, and surveillance, will be highlighted in the chapter through the references from the installation and evaluation processes. The work presents a viable approach for object detection and classification in changeful visual arenas.

Diabetes presents significant health risks globally, necessitating precise blood glucose monitoring to prevent serious repercussions including blindness, renal illness, kidney failure, heart disease, and even death from hyperglycemia or hypoglycemia, it is imperative to maintain normal blood glucose levels. However, regular blood glucose monitoring can be difficult for diabetics, and current non-invasive techniques sometimes do not assess blood sugar levels accurately or directly. In order to solve this problem, this study suggests a wearable optical system that is affordable and low-complexity. In this study, a wearable optical system has been proposed which can address the challenges in the accuracy and convenience in existing methods. This system used an Arduino Nano as a central control unit and a laser-transmitted module for blood glucose measurement. Light Dependent Resistors (LDRs) is used to detect and measure the intensity of laser light passing through the skin and impressed by blood glucose levels. The results are displayed on Organic Light Emitting Diode (OLED). During one weak trial, the system achieved average error present of 7.6% and 3.9% for before and after meal blood glucose concentration. The aim of this study is to enhance the lifestyle of diabetic patients by providing user-friendly technology for convenient blood glucose monitoring. It focuses on the potential benefits of non-invasive approaches and concentrates on the importance of the proposed wearable optical system in improving healthcare outcomes.