Brain MR images are the most suitable method for detecting chronic nerve diseases such as brain tumors, strokes, dementia, and multiple sclerosis. They are also used as the most sensitive method in evaluating diseases of the pituitary gland, brain vessels, eye, and inner ear organs. Many medical image analysis methods based on deep learning techniques have been proposed for health monitoring and diagnosis from brain MRI images. CNNs (Convolutional Neural Networks) are a sub-branch of deep learning and are often used to analyze visual information. Common uses include image and video recognition, suggestive systems, image classification, medical image analysis, and natural language processing. In this study, a new modular deep learning model was created to retain the existing advantages of known transfer learning methods (DenseNet, VGG16, and basic CNN architectures) in the classification process of MR images and eliminate their disadvantages. Open-source brain tumor images taken from the Kaggle database were used. For the training of the model, two types of splitting were utilized. First, 80% of the MRI image dataset was used in the training phase and 20% in the testing phase. Secondly, 10-fold cross-validation was used. When the proposed deep learning model and other known transfer learning methods were tested on the same MRI dataset, an improvement in classification performance was obtained, but an increase in processing time was observed.

Monkeypox disease is caused by a virus that causes lesions on the skin and has been observed on the African continent in the past years. The fatal consequences caused by virus infections after the COVID pandemic have caused fear and panic among the public. As a result of COVID reaching the pandemic dimension, the development and implementation of rapid detection methods have become important. In this context, our study aims to detect monkeypox disease in case of a possible pandemic through skin lesions with deep-learning methods in a fast and safe way. Deep-learning methods were supported with transfer learning tools and hyperparameter optimization was provided. In the CNN structure, a hybrid function learning model was developed by customizing the transfer learning model together with hyperparameters. Implemented on the custom model MobileNetV3-s, EfficientNetV2, ResNET50, Vgg19, DenseNet121, and Xception models. In our study, AUC, accuracy, recall, loss, and F1-score metrics were used for evaluation and comparison. The optimized hybrid MobileNetV3-s model achieved the best score, with an average F1-score of 0.98, AUC of 0.99, accuracy of 0.96, and recall of 0.97. In this study, convolutional neural networks were used in conjunction with optimization of hyperparameters and a customized hybrid function transfer learning model to achieve striking results when a custom CNN model was developed. The custom CNN model design we have proposed is proof of how successfully and quickly the deep learning methods can achieve results in classification and discrimination.

Human action recognition (HAR), deeply rooted in computer vision, video surveillance, automated observation, and human computer interaction (HCI), enables precise identification of human actions. Numerous research groups have dedicated their efforts to various applications and problem domains in HAR systems. They trained classification models using diverse datasets, enhanced hardware capabilities, and employed different metrics to assess performance. Although several surveys and review articles have been published periodically to highlight research advancements in HAR, there is currently no comprehensive and up-to-date study that encompasses architecture, application areas, techniques/algorithms, and evaluation methods as well as challenges and issues. To bridge this gap in the literature, this article presents a comprehensive analysis of the current state of HAR systems by thoroughly examining a meticulously chosen collection of 136 publications published within the past two decades. These findings have significant implications for researchers engaged in different aspects of HAR systems.

A massive volume of unstructured data in the form of comments, opinions, and other sorts of data is generated in real-time with the growth of web 2.0. Due to the unstructured nature of the data, building an accurate predictive model for sentiment analysis remains a challenging task. While various DNN architectures have been applied to sentiment analysis with encouraging results, they suffer from high dimensional feature space and consider various features equally. State-of-the-art methods cannot properly leverage semantic and sentiment knowledge to extract meaningful relevant contextual sentiment features.This paper proposes a sentiment and context-aware hybrid DNN model with an attention mechanism that {intelligently learns and highlights salient features of relevant sentiment context in the text. We first use integrated wide coverage sentiment lexicons to identify text sentiment features then leverage bidirectional encoder representation from transformers to produce sentiment-enhanced word embeddings for text semantic extraction. After that, the proposed approach adapts the BiLSTM to capture both word order/contextual text semantic information and the long-dependency relation in the word sequence. Our model also employs an attention mechanism to assign weight to features and give greater significance to salient features in the word sequence. Finally, CNN is utilized to reduce the dimensionality of feature space and extract the local key features for sentiment analysis. The effectiveness of the proposed model is evaluated on real-world benchmark datasets demonstrating that the proposed model significantly improves the accuracy of existing text sentiment classification.

Recurrent Neural Networks (RNNs) and their variants have been demonstrated tremendous successes in modeling sequential data such as audio processing, video processing, time series analysis, and text mining. Inspired by these facts, we propose human activity recognition technique to proceed visual data via utilizing convolution neural network (CNN) and Bidirectional-gated recurrent unit (Bi-GRU). Firstly, we extract deep features from frames sequence of human activities videos using CNN and then select most important features from the deep appearances to improve performance and decrease computational complexity of the model. Secondly, to learn temporal motions of frames sequence, we design Bi-GRU and feed those deep-important features extracted from frames sequence of human activities to Bi-GRU which learn temporal dynamics in forward and backward direction at each time step. We conduct extensive experiments on realistic videos of human activity recognition datasets YouTube11, HMDB51 and UCF101. Lastly, we compare the obtained results with existing methods to show the competence of our proposed technique.

This review presents a detailed examination of the most recent advancements in positron emission tomography–computed tomography (PET-CT) multimodal imaging over the past five years. The fusion of PET and CT technologies has revolutionized medical imaging, offering unprecedented insights into both anatomical structure and functional processes. The analysis delves into key technological innovations, including advancements in image reconstruction, data-driven gating, and time-of-flight capabilities, highlighting their impact on enhancing diagnostic accuracy and clinical outcomes. Illustrative case studies underscore the transformative role of PET-CT in lesion detection, disease characterization, and treatment response evaluation. Additionally, the review explores future prospects and challenges in PET-CT, advocating for the integration and evaluation of emerging technologies to improve patient care. This comprehensive synthesis aims to equip healthcare professionals, researchers, and industry stakeholders with the knowledge and tools necessary to navigate the evolving landscape of PET-CT multimodal imaging.