The rapid proliferation of Internet of Medical Things (IoMT) devices has created unprecedented demands for reliable, low-latency, and high-bandwidth communication networks in healthcare environments. Traditional networking approaches struggle to meet the diverse quality of service requirements of medical applications, ranging from real-time patient monitoring to large-scale medical imaging transfers. Intent-based networking (IBN) emerges as a promising paradigm that translates high-level healthcare network intentions into automated network policies, enabling self-managing and self-optimizing networks. This paper presents a novel intelligent routing framework for IoMT communications based on IBN principles, utilizing a dueling double deep Q-network (DDQN) approach to optimize network resource allocation and routing decisions. The proposed method addresses the multi-objective optimization problem of simultaneously satisfying bandwidth, latency, and reliability requirements for heterogeneous medical traffic flows. We model the IBN routing problem as a reinforcement learning task and develop a dueling DDQN algorithm that learns optimal routing policies through interaction with a network environment model. Comparative analysis against baseline methods reveals 12-42% improvements in medical intent satisfaction across application categories. The framework exhibits robust convergence and maintains advantages across network scales from clinics to enterprise systems. Sensitivity analysis validates optimal bandwidth discretization, balancing efficiency with clinical requirements. Results confirm practical applicability for IoMT deployments where network reliability impacts patient care.

Integrating medical Internet of Things (IoT) systems with 6G networks presents unprecedented real-time medical image analysis opportunities. However, it introduces significant challenges when models trained on one device fail to generalize to images from different equipment. This paper proposes enhanced target domain representation for cross-domain medical image segmentation (E-TDRCMIS), a novel high-definition CT image segmentation approach in 6G medical IoT environments. Our method addresses domain shift challenges through a three-component architecture optimized for distributed healthcare systems. Our shared feature learning module employs adversarial techniques with gradient reversal layers to extract domain-invariant features while minimizing computational overhead at network edges. The shared feature enhancement module implements a multi-level consistency regularization strategy that compares predictions from shallow and deep features to strengthen feature representation without increasing bandwidth requirements. Finally, the target domain generalization module reconstructs complete feature representations. Experiments on chest and spine structure datasets demonstrate that E-TDRCMIS significantly outperforms existing methods in cross-domain high-definition CT image segmentation. Specifically, E-TDRCMIS achieved Dice scores of 87.9% and 91.3% on the LUNA16 and VerSe2020 datasets, respectively, with corresponding average symmetric surface distance (ASD) values of 1.2mm and 0.9mm. This represents improvements of 3.0% (Dice) and 20.0% (ASD) over the next best performing method (MRLA-Net) on LUNA16, and 2.5% (Dice) and 30.8% (ASD) improvements on VerSe2020.

Healthcare supply chains manage increasing volumes of high-resolution computed tomography (CT) scans daily. The scan of medical images uses storage infrastructure increases transmission costs and elevates carbon emissions. Traditional compression methods achieve limited ratios for medical images, while current deep-learning approaches require intensive computational resources, making them unsuitable for sustainable healthcare operations. This study introduces ESGC-Net, an Artificial Intelligence of Medical Things (AIoMT) driven framework that optimizes secure medical image distribution throughout the healthcare supply chain while reducing environmental impact. Our approach integrates deep spatiotemporal learning with subset-based coding to create an environmentally sustainable compression pipeline. The framework processes CT image sequences using regional healthcare edge devices, enabling efficient data flow from imaging centers to cloud storage and clinical access points. The AIoMT integration allows for distributed processing at multiple points in the supply chain, reducing data transfer volumes while maintaining data security through adaptive encryption that protects patient privacy across all supply chain nodes. Experimental results demonstrate that ESGC-Net attains a 4.85 compression ratio while preserving diagnostic image quality, reducing processing time by 41.0% and energy consumption by 31.5% compared to existing methods. The subset-optimal encoding technique decreases encoding time by 47.8%, enabling faster image transmission between healthcare facilities.

Smart medical rescue vehicles (SMRVs) are crucial in providing timely and effective emergency medical services in urban environments. However, the efficiency and safety of SMRVs are often hindered by dynamic and complex traffic conditions, leading to longer response times and increased risk of accidents. To address these challenges, this article proposes a novel lane-changing strategy called AIoT-LC, which leverages the Augmented Intelligence of Things (AIoT) framework to enable efficient vehicle road cooperation for smart medical rescue. First, a deep Q-network is employed to process real-time data streaming from SMRVs and support their collaborative management with roads and pedestrians. Then, a cooperative lane-changing strategy is developed to ensure the safe and efficient navigation of SMRVs through dynamic traffic environments, considering factors, such as safety distance, lane-changing trajectory planning, and multivehicle coordination. Finally, a collision resolution and avoidance strategy is proposed for SMRVs at intersections, leveraging the vehicle-infrastructure cooperative environment and the AIoT framework to minimize the risk of collisions and optimize the intersection crossing process. The experimental results demonstrate that the proposed AIoT-LC method significantly outperforms existing approaches, reducing average travel time for SMRVs by up to 25%, improving average speed by 15%, and increasing the success rate of emergency responses by 20% points compared to the best performing baseline method. Additionally, the proposed method reduces fuel consumption by 11.7% and CO2 emissions by 11.7% while improving overall traffic flow efficiency by 8.9%.

Modern vehicles involve large number of sensors, cameras and communication systems for real-time traffic management, collision avoidance and vehicle health monitoring. As the Internet of Vehicles (IoV) ecosystem evolve with more number of connected vehicles, handling of the enormous data is a challenge. This is overcome with the promising distributed edge intelligence (DEI) approach in which the computational tasks are distributed among the intelligent road side units (RSUs) at the network edge. The edge servers cooperate among themselves so as not to overload the central cloud server. This article presents a cooperative vehicular communication network which exploits the existing 5G infrastructure in roadside building as the edge/relay nodes. To overcome the communication and energy overhead, network resource management is enabled through proposed edge node selection algorithm. Further, a joint edge node and antenna selection algorithm is proposed for enhanced energy efficiency (EE) and reduced outage. The closed-form expression for the outage probability of the proposed cooperative communication scheme is derived. Our analysis shows that the proposed selection approach achieves improved outage probability and energy-efficiency. In particular, the proposed edge node selection approach improves the EE by 10.46% at total transmit power to noise power ratio of 16 dB. Moreover, the overall system performance is further enhanced by proposing a joint selection scheme. Specifically, the analysis shows that the energy-efficiency improves by 27.87% with the joint selection scheme. In the end, a use case scenario of DEI empowered IoVs in emergency healthcare support is discussed along with the future research directions.

In healthcare environments, accurate and real-time patient localization and monitoring are crucial for ensuring patient safety and improving operational efficiency. This article proposes DeepSense-Healthcare, a hybrid CNN-LSTM framework integrated with adaptive resource management to enhance near-field (NF) IoT-based patient localization and monitoring in healthcare facilities. By combining convolutional neural networks (CNNs) for spatial feature extraction with long short-term memory (LSTM) networks for temporal modeling, DeepSense-Healthcare captures complex spatial-temporal patterns in NF signals, achieving high localization accuracy. An adaptive resource management module is incorporated to optimize computational load, dynamically adjusting resource allocation based on patient activity levels, thereby improving energy efficiency and maintaining responsiveness. We evaluate the proposed framework against baseline models through extensive experiments on the SEED-VIG and ILM datasets across various activity levels. The results demonstrate that DeepSense-Healthcare outperforms conventional methods in localization accuracy, energy efficiency, and latency during various activity scenarios. These findings underscore the effectiveness of DeepSense-Healthcare as a robust and efficient solution for continuous patient monitoring in dynamic healthcare settings.

In the realm of medical data processing, particularly in the diagnosis and monitoring of cardiac diseases, the analysis of electrocardiogram (ECG) signals represents a critical challenge, especially with the burgeoning volume of ECG Big Data. Traditional methods and existing research often fall short in effectively analyzing this data, limited by their inability to fully capture the complex and nonlinear patterns inherent in ECG signals. Addressing these limitations, in this article, we introduce a novel deep neuro-fuzzy model augmented with multimodal feature fusion. Our method ingeniously combines the power of neuro-fuzzy systems with the robust feature extraction capabilities of deep learning, specifically leveraging a Transformer-based architecture, to analyze both ECG signals and their corresponding spectral images. This multimodal fusion not only enriches the model's input data, providing a comprehensive understanding of cardiac signals, but also enhances the adaptability and accuracy of cardiac arrhythmia detection. We rigorously validate our approach on the MIT-BIH arrhythmia database, conducting a series of experiments, including performance evaluations and ablation studies, to highlight the significant contributions of the multimodal feature fusion and neuro-fuzzy module. The results achieve significant improvements in classification metrics: an accuracy of 98.46% and an F1 score of 99.1%. Moreover, we benchmark the Transformer's feature extraction performance against other architectures, such as ResNet. The results unequivocally demonstrate our model's superiority and illustrate the potential of integrated neuro-fuzzy and deep learning approaches in overcoming the current limitations of ECG signal analysis.

In medical diagnostics, the accurate classification and analysis of biomedical signals play a crucial role, particularly in the diagnosis of neurological disorders such as epilepsy. Electroencephalogram (EEG) signals, which represent the electrical activity of the brain, are fundamental in identifying epileptic seizures. However, challenges such as data scarcity and imbalance significantly hinder the development of robust diagnostic models. Addressing these challenges, in this paper, we explore enhancing medical signal processing and diagnosis, with a focus on epilepsy classification through EEG signals, by harnessing AI-generated content techniques. We introduce a novel framework that utilizes generative adversarial networks for the generation of synthetic EEG signals to augment existing datasets, thereby mitigating issues of data scarcity and imbalance. Furthermore, we incorporate an attention-based temporal convolutional network model to efficiently process and classify EEG signals by emphasizing salient features crucial for accurate diagnosis. Our comprehensive evaluation, including rigorous ablation studies, is conducted on the widely recognized Bonn Epilepsy Data. The results achieves an accuracy of 98.89% and F1 score of 98.91%. The findings demonstrate substantial improvements in epilepsy classification accuracy, showcasing the potential of AI-generated content in advancing the field of medical signal processing and diagnosis.

The rapid advancement of large language models (LLMs) has opened up new possibilities for transforming healthcare practices, patient interactions, and medical report generation. This paper explores the application of LLMs in developing medical chatbots and virtual assistants that prioritize clinical accuracy. We propose a novel multi-turn dialogue model, including adjusting the position of layer normalization to improve training stability and convergence, employing a contextual sliding window reply prediction task to capture fine-grained local context, and developing a local critical information distillation mechanism to extract and emphasize the most relevant information. These components are integrated into a multi-turn dialogue model that generates coherent and clinically accurate responses. Experiments on the MIMIC-III and n2c2 datasets demonstrate the superiority of the proposed model over state-of-the-art baselines, achieving significant improvements in perplexity, BLEU-2, recall at K scores, medical entity recognition, and response coherence. The proposed model represents a significant step in developing reliable and contextually relevant multi-turn medical dialogue systems that can assist patients and healthcare professionals.

Driver fatigue detection is a critical challenge in Transportation Cyber-Physical Systems (T-CPS), where existing methods often face significant limitations. Conventional approaches typically struggle with issues such as limited accuracy, slow convergence, and high computational costs. These methods often fail to capture the complex temporal and spatial patterns inherent in multimodal physiological signals, such as EEG and EOG, leading to suboptimal performance in real-world scenarios. To address these challenges, we propose a novel approach that leverages Depthwise Separable Convolutional Neural Networks (DSCNNs) for driver fatigue detection. Our method integrates electroencephalogram (EEG) and electrooculogram (EOG) signals through an innovative feature extraction and fusion process, enhancing the system’s ability to detect fatigue with high accuracy and efficiency. The DSCNN model is designed to overcome the limitations of traditional methods by utilizing depthwise separable convolutions, which reduce computational complexity while maintaining robust performance. We validate our approach using the SEED-VIG dataset and the NTHU drowsy driver dataset, which encompasses a range of driving conditions. The DSCNN model outperforms conventional models in both accuracy and computational efficiency. Specifically, DSCNN achieves the highest F1 score and the lowest time per epoch, making it highly suitable for real-time applications in T-CPS. This advancement represents a significant improvement in driver safety by providing more timely and reliable fatigue warnings, thus advancing the capabilities of intelligent transportation systems.

Conventional security architectures and models are considered single-network architecture solutions, which assume that devices authenticated within the network are implicitly trusted. However, such an approach is unsuitable for next-generation networks (NGNs). Zero-trust security was introduced to overcome these challenges using context-aware, dynamic, and intelligent authentication schemes. This paper proposes a novel zero-trust blockchain-enabled framework for secure next-generation healthcare communication network (HCN). The proposed framework integrates zero-trust and blockchain to provide a decentralized, secure, and intelligent solution for healthcare communication in NGNs. The system model comprises three components: HCN user identity modeling, blockchain and risk assessment-based access control, and dynamic trust gateway. The user identity modeling component utilizes attribute-based user behavior trajectory features, while the access control component leverages smart contracts-based risk assessment. The dynamic trust gateway component employs a consensus mechanism to achieve dynamic gateway switching and enhance network resilience. Simulation results demonstrate that the proposed framework achieves 31% lower calculation delays, 3% higher trust values, and 3% better attack detection accuracy compared to best baseline methods. It also exhibits a 2% improvement in access control granularity and maintains 95% network throughput under various failure scenarios.