Recent Advances in Computer Science and Communications - Volume 17, Issue 3, 2024

Introduction: To propose a medical image registration method with significant performance improvement. The spatial transformation obtained by the traditional deformable image registration technology is not smooth enough, and the calculation amount is too large to solve the optimization parameters. The network model proposed based on deep learning medical image registration technology has some limitations, which cannot guarantee the registration of topological structures, resulting in the loss of spatial features. It makes the model have topological conservation and transform reversibility, has the ability to learn more multi-scale features and complex image structures, and captures finer changes while clearly encoding global correlation. Method: Based on the core UNet model, a deformable image registration method with a new architecture Broad-UNet-diff is proposed. The model is equipped with asymmetric parallel convolution and uses diffeomorphism mapping. Result: Compared with the seven classical registration methods under the brain MRI datasets, the proposed method has significantly improved the registration performance. In particular, compared with the advanced TransMorph-diff registration method, the Dice score can be improved by 12 %, but only the 1/10 parameters are needed. Conclusion: This method confirms its convincing effectiveness and accuracy.

Background: A new DWT-ERT-based fault location method is suggested in the IEEE test feeder. Objective: The fault location approach in the distribution network has been proposed in this paper that utilizes the discrete wavelet transform (DWT) and ensemble regression tree (ERT). Methods: The fault location methodology has been validated by simulations conducted on an IEEE 13 bus node test feeder. Results: The results show that the suggested solution has low compute burden and memory requirements, and is unaffected by system and fault situations. Conclusion: In this study, the fault location approach for the distribution system employing DWT and ERT has been proposed.

Background: Pneumonia is one of the leading causes of death and disability due to respiratory infections. The key to successful treatment of pneumonia is in its early diagnosis and correct classification. PneumoniaNet is a unique deep-learning model based on CNN for identifying pneumonia on chest X-rays. Objective: A deep learning model that combines convolutional, pooling, and fully connected layers is presented in this study. Methods: In order to learn how to identify cases of pneumonia and healthy controls on chest X-ray pictures, PneumoniaNet was trained on a large labeled library of such images. A robust data augmentation technique was adopted to enhance the model generalization and training set diversity. Standard measures like as accuracy, precision, recall, and F1-score were applied to PneumoniaNet's performance evaluation. Results: The suggested model performed effectively in detecting pneumonia cases with an accuracy of 93.88%. Conclusion: The model was evaluated against the current state-of-art methods and showed that PneumoniaNet outperformed the other models.

Background: Cross-domain Sentiment Classification is a well-researched field in sentiment analysis. The biggest challenge in CDSC arises from the differences in domains and features, which cause a decrease in model performance when applying source domain features to predict sentiment in the target domain. To address this challenge, several feature selection methods can be employed to identify the most relevant features for training and testing in CDSC. Methods: The primary objective of this study is to perform a comparative analysis of different feature selection methods on the various CDSC tasks. In this study, statistical test-based feature selection methods using 18 classifiers for the CDSC task has been implemented. The impact of these feature selection methods on Amazon product reviews, specifically those in the DVD, Electronics, Kitchen, and TV domains, has been compared. Total 12x18 experiments were conducted for each feature selection method by varying source and target domain pairs from the Amazon product reviews dataset and by using 18 classifiers. Performance evaluation measures are accuracy and f-score. Results: From the experiments, it has been inferred that the CSDC task depends on various factors for a good performance, from the right domain selection to the right feature selection method. We have concluded that the best training dataset is Electronics as it gives more precise results while testing in either domain selected for our study. Conclusion: Cross-domain sentiment analysis is a dynamic and interdisciplinary field that offers valuable insights for understanding how sentiment varies across different domains.

Background: Wireless Sensor Networks (WSNs) have gained significant attention due to their diverse applications, including border area security, earthquake detection, and fire detection. WSNs utilize compact sensors to detect environmental events and transmit data to a Base Station (BS) for analysis. Energy consumption during data transmission is a critical issue, which has led to the exploration of additional energy-saving techniques, such as clustering. Objective: The primary objective is to propose an algorithm that selects optimal Cluster Heads (CHs) through a fuzzy-based genetic approach. This algorithm aims to address energy consumption concerns, enhance load balancing, and improve routing efficiency within WSNs. Methods: The proposed algorithm employs a fuzzy-based genetic approach to optimize the selection of CHs for data transmission. Four key parameters are considered: the average remaining energy of CHs, the average distance between CHs and the BS, the average distance between member nodes and CHs, and the standard deviation of the distance between member nodes and CHs. Results: The algorithm's effectiveness is demonstrated through simulation results. When compared to popular models like LEACH, MOEES, and FEEC, it demonstrates an 8-20% improvement in the lifetime of WSNs. The proposed approach achieves enhanced efficiency, lifetime extension, and improved performance in CH selection, load balancing, and routing. Conclusion: In conclusion, this study introduces a novel algorithm that utilizes fuzzy-based genetic techniques to optimize CH selection in WSNs. By considering four key parameters and addressing energy consumption challenges, the proposed algorithm offers significant improvements in efficiency, lifespan, and overall network performance, as validated through simulation results.

Background: Feature selection (FS) is critical for high dimensional data analysis. Ensemble based feature selection (EFS) is a commonly used approach to develop FS techniques. Rank aggregation (RA) is an essential step in EFS where results from multiple models are pooled to estimate feature importance. However, the literature primarily relies on static rule-based methods to perform this step which may not always provide an optimal feature set. The objective of this study is to improve the EFS performance using dynamic learning in RA step. Method: This study proposes a novel Supervised Rank Aggregation (SRA) approach to allow RA step to dynamically learn and adapt the model aggregation rules to obtain feature importance. Results: We evaluate the performance of the algorithm using simulation studies and implement it into real research studies, and compare its performance with various existing RA methods. The proposed SRA method provides better or at par performance in terms of feature selection and predictive performance of the model compared to existing methods. Conclusion: SRA method provides an alternative to the existing approaches of RA for EFS. While the current study is limited to the continuous cross-sectional outcome, other endpoints such as longitudinal, categorical, and time-to-event data could also be used.

Background: Human physical activity recognition is challenging in various research eras, such as healthcare, surveillance, senior monitoring, athletics, and rehabilitation. The use of various sensors has attracted outstanding research attention due to the implementation of machine learning and deep learning approaches. Aim: This paper proposes a unique deep learning framework based on motion signals to recognize human activity to handle these constraints and challenges through deep learning (e.g., enhance CNN, LR, RF, DT, KNN, and SVM) approaches. Method: This research article uses the BML (Biological Motion Library) dataset gathered from thirty volunteers with four various activities to analyze the performance metrics. It compares the evaluated results with existing results, which are found by machine learning and deep learning methods to identify human activity. Result: This framework was successfully investigated with the help of laboratory metrics with convolutional neural networks (CNN) and achieved 89.0% accuracy compared to machine learning methods. Conclusion: The novel work of this research is to increase classification accuracy with a lower error rate and faster execution. Moreover, it introduces a novel approach to human activity recognition in the BML dataset using the CNN with Adam optimizer approach.