ABSTRACTWe address the hidden tumour visualization problem in augmented monocular liver laparoscopy. Conveying a hidden tumour's depth correctly to the surgeon in augmented monocular laparoscopy is extremely difficult and still forms an unsolved problem. The depth conveyance can be split into two subsequent problems. First, designing a visualization that convinces the user to see the tumour inside the organ. Second, enhancing this visualization so that it also provides metric depth perception. The most promising visualization methods rely on a preoperative CT organ model with the tumour to be registered to an intraoperative laparoscopic image. Such a registration allows the organ's intraoperative shape mesh to be overlaid on top of the augmented tumour. The overlaid organ mesh guarantees a partial occlusion on the augmented tumour. This provides a powerful depth cue for the surgeon's perception. However, this type of registration, especially in liver laparoscopy, is usually not real‐time and sometimes not possible. This is because of the liver deformation and lack of matchable features between the multimodal images. Subsequently, the tumour augmentation cannot be carried out continuously to guide the surgeon. We propose a novel visualization method to address these limitations. The proposed method replaces the deformable preoperative to intraoperative liver registration with a rigid tumour registration via laparoscopic ultrasound imaging. The proposed method handles surgical tool occlusions, runs faster, and outperforms the state of the art in terms of depth perception, as shown in the user study.

Photoplethysmography (PPG) enables non-invasive estimation of vital signs across a range of sensors, from wrist wearables and clinical monitors to smartphone cameras. We present parameter-efficient, end-to-end deep models for heart rate (HR), blood oxygen saturation ( ) and respiratory rate (RR) estimation. These models operate directly on PPG signals and integrate preprocessing as a layer. Our family of fully convolutional architectures includes a baseline FCN, a residual FCN, a ConvNeXt-inspired FCN and a compact DCT-based variant. These models achieve strong cross-dataset performance while using only 3.3K-53K parameters. On PPG-DaLiA, the best performing model achieves 6.07 2.70 bpm MAE for HR. In HR, and RR tasks on the BIDMC dataset, the best-performing method achieves error rates of 2.2 bpm, 3.14% and 1.49 breaths per minute, respectively. We also release MTHS, a smartphone-based vital-sign dataset comprising 62 fingertip-PPG recordings (35 men and 27 women) with 1-Hz HR and annotations. Additionally, we demonstrate practical feasibility via on-device deployment using TFLite on smartphones. These results show that compact, fully convolutional and DCT-based designs provide accurate, generalizable, and deployment-ready PPG vital-sign estimators across sensors anddatasets.

Early diagnosis of thoracic diseases using chest x-ray imaging remains a critical challenge, particularly in resource-constrained healthcare environments where data sharing is restricted due to privacy concerns. Federated learning (FL) offers a decentralized solution by enabling collaborative model training without sharing sensitive patient data. However, integrating privacy-preserving mechanisms such as differential privacy (DP) introduces additional challenges related to performance degradation and computational overhead. In this study, we present a unified FL framework for multi-label thoracic disease classification using multiple convolutional neural network (CNN) architectures, including ResNet50, DenseNet169, EfficientNet variants and MobileNetV3. Unlike prior studies focusing on single-model evaluation, this work provides a controlled comparative analysis under identical FL settings and investigates the impact of client scalability (5-10 clients) on model performance. Furthermore, we conduct a comprehensive empirical analysis of the privacy utility trade-off by integrating DP with varying privacy budgets (ε=1, 15 and 30). Experimental results on the CheXpert and NIH Chest x-ray14 datasets demonstrate that the proposed EfficientNet-B3-based federated model achieves a mean AUC of 0.8027, while maintaining robustness across decentralized settings. The integration of DP leads to a predictable reduction in performance, with mean AUC ranging from 0.60 to 0.64, highlighting the inherent trade-off between privacy and diagnostic accuracy. The findings emphasize the practical viability of FL for privacy-sensitive medical imaging applications and provide insights into model selection, scalability and privacy configuration for real-world deployment. The source code for this study is publicly accessible at https://github.com/Zulqarnain8-8/FEDERATED_LEARNING_FOR_THORACIC_DISEASE_CLASSIFICATION.

Telemedicine has been introduced as a new and effective method in dealing with public health challenges, improving access to health care, and reducing healthcare costs in today's world. In order to solve the problems of a single access network and the unknown network status of the telemedicine system in complex environments, a multi-link aggregation algorithm based on link priority (MLA-LP) was proposed. Firstly, the network parameters, such as packet loss rate (PLR) and round-trip time (RTT), of each link were obtained, and then MLA-LP was realized by using the congestion control algorithm based on sender and the traffic allocation algorithm based on link priority. Finally, the MLA-LP algorithm was tested on the test platform based on HoloWAN 1200 network simulator. The comparison algorithm was the multilink aggregation algorithm based on the weightedround-robin algorithm (MLA-WRR). MLA-LP algorithm has higher uplink throughput than MLA-WRR under the same link PLR and delay. When the link is suddenly abnormal (burst packet loss, delay, etc.), the MLA-LP algorithm could quickly avoid the abnormal link and transfer traffic within 2 s; In the mobile scenario, compared with a single network and MLA-WRR algorithm, MLA-LP algorithm has the best video transmission effect, with an average delay of 136 ms, an average bit rate of 2.31 Mbps, an average PLR of 0.04%, and video jamming times close to 0. The MLA-LP algorithm could effectively aggregate link bandwidth and quickly avoid abnormal links, ensuring the robustness of the access network.In order to solve the problems of a single access network and the unknown network status of telemedicine system in complex environment, a MLA-LP was proposed. MLA-LP algorithm has higher uplink throughput and could quickly avoid the abnormal link and transfer traffic within 2 s, ensuring the robustness of the access network.

A substantial amount of literature has been published on epileptic seizures. However, adequate evidence is still lacking to demonstrate that utilising explainable artificial intelligence for epileptic seizures can ensure an individual's safety. Furthermore, there is a need to define the fundamental challenges and opportunities present in the current state-of-the-art solutions and guide efforts towards responsible artificial intelligence. To identify fundamental challenges and opportunities in the existing state-of-the-art solutions available for explainable artificial intelligence-based epileptic seizure onset early warning: towards responsible artificial intelligence. The PRISMA checklist was utilised to develop this report. Papers were extracted from original articles and prior conference studies published in reputable databases such as PubMed, IEEE Xplore, ScienceDirect, Scopus and Google Scholar from January 2019 to 17 November 2024. Rayyan's online platform was used to identify duplicates, inclusions and exclusions of papers. This systematic review protocol was registered with the PROSPERO database. The included papers were assessed based on Microsoft's Responsible Artificial Intelligence template. The Responsible AI Impact Assessment Template, Principle 3 (transparency and explainability), determined a high-risk rating. A total of 26 studies are included based on the established inclusion and exclusion criteria. This study investigated 14.29% of responsible artificial intelligence principles applied in at least one paper with a high-risk rate. The results indicate that to transform researched solutions into practical applications, epileptic monitoring applications should be tested within the eight principles set by Microsoft. The black box explanation lacks insight into the deep internal features and operational methods, suggesting that further investigation is necessary. Systematic Review Registration ID: CRD42024544.

With the transformation of modern medical care towards a prevention-centred model and the growing demand for high-density sensor deployments in wireless body area networks (WBANs), traditional orthogonal multiple access (OMA) schemes face critical limitations in spectral efficiency and energy sustainability. Non-orthogonal multiple access (NOMA) combined with radio frequency energy harvesting (RFEH) emerges as a promising solution, yet existing research often overlooks the strict energy causality constraints of sensors. To address this issue, this paper proposes a novel joint optimization framework for NOMA-assisted WBANs, which strategically pairs predefined sensor groups (classified by body positions) into NOMA clusters and optimizes power allocation while accounting for harvested RF energy. The proposed sorted-greedy joint algorithm enhances successive interference cancellation (SIC) efficiency by pairing sensor groups with the largest channel gain differences, followed by a greedy power allocation strategy that maximizes throughput under energy and power constraints. Simulation results demonstrate that the proposed algorithm outperforms conventional greedy and random clustering approaches in total system throughput across varying numbers of sensor groups, sensors per group, and transmission powerlevels.

ABSTRACTPeripheral intravenous catheter (PIVC) insertion is a common yet challenging procedure. Although ultrasound guidance improves procedural accuracy and patient outcome, its complexity limits its routine adoption to highly experienced clinicians. This paper introduces a virtual reality (VR) simulator developed specifically for training in ultrasound‐guided PIVC insertions. This study aims to validate the simulator's realism and relevance through face, content, and construct assessments, and to demonstrate its utility as a platform for comparing various approaches to PIVC insertion. Thirty participants from diverse medical backgrounds and levels of expertise completed three scenarios, each featuring a different procedural technique, within the simulator's controlled virtual environment. The simulator demonstrated strong face and content validity, with participants rating its realism at 7.1/10 and enjoyment at 8.2/10. Performance data showed that expert participants maintained higher success rates and performance across all procedural scenarios, supporting the simulator's construct validity. In the standard approach scenario, novices required 230.91 ± 158.77 s to complete the task and achieved only a 45% success rate compared to experts' 95.48 ± 65.74 s and 80% success rate. In the procedural scenario involving an alignment assistance device, where needle insertion was aligned with the ultrasound image plane, novice success rates increased to 75% and the number of attempts decreased from 8.95 ± 6.69 to 2.75 ± 2.67, narrowing the performance gap with experts. These findings highlight the simulator's potential not only as an effective training tool but also as a platform for the objective evaluation of different procedural techniques.

ABSTRACT In this paper, we introduce SAMmorph, a novel multimodal medical image registration framework addressing challenges in aligning cross‐modality images due to significant appearance differences and heavy reliance on expensive annotations. It comprises three key components: the VoxelSAM module for generating pseudo‐labels, the FusionLayer which takes moving/fixed images and their pseudo‐labels as input, integrates them via two optional schemes (stack: concatenating along the channel dimension to form a 4‐channel tensor; enhance: modulating image intensity with mask confidence to generate a 3‐channel tensor) to produce combined features for registration, and the deformable registration network which includes two architectures (joint and independent registration networks) to implement the bidirectional registration strategy. Experiments demonstrate that SAMmorph outperforms state‐of‐the‐art methods across various metrics on the ‐RegPro dataset, achieving a dice similarity coefficient (DSC) of 0.906, target registration error (TRE) of 6.06 mm, compared to those being 0.789 and 7.632 mm achieved by the strongest tested baseline method CorrMLP, with statistical significance confirmed by Wilcoxon signed‐rank tests (). This work presents a promising approach for improving medical image registration accuracy and efficiency, with potential applications in diverse medical imaging scenarios. The source codes have been released at: https://github.com/qilin‐world/SAMmorph .

This study aimed to analyse the effects of co-contraction on quadriceps and hamstring muscles during sit-to-stand (STS) tasks for females wearing shoes with different heel heights. The study aimed to identify compensatory strategies during the STS tasks in response to excessive muscle co-contraction during high-heeled gait. Sixteen healthy young women (age: 24.4±1.7 years, body mass index: 18.4±1kg/m2, weight: 50.2±5.2kg, height: 1.63±4.4m) participated in this study. Electromyography signals were recorded from three quadriceps (vastus medialis, vastus lateralis, and rectus femoris) and one hamstring (semitendinosus) muscles. The participants wore shoes with different heights, including 4, 6, 8, and 10cm. For each heel height, the co-contraction index is computed to measure postural balance using the quadriceps to hamstring muscle pairs. The results that were obtained and quantified with statistical measures show that for elevated shoes, if co-contraction increases, both quadriceps and hamstring muscles tend to compensate. This suggests that the capacity of the quadriceps and hamstring muscles to compensate is essential to retain normal walking and STS tasks in co-contracted persons. However, the compensation mechanisms may induce imbalance, muscle stiffness, and fatigue for women who regularly use high-heeled shoes during sit-to-stand tasks.

Sepsis, a life-threatening condition with high mortality and readmission rates, demands precise and timely management to improve patient outcomes. Despite advancements, identifying the critical activities within sepsis treatment pathways remains a challenge, limiting the effectiveness of interventions. This study addresses this issue by utilising comparative process mining techniques to analyse sepsis trajectories, focusing on key performance metrics-sojourn time, arrival rate and finish rate-across distinct patient clusters. The analysis is based on real-life event logs from a hospital's sepsis cases, using K-means clustering to segment patients by age, severity and key clinical indicators. The study reveals critical activities such as 'Return ER', 'Admission IC', and 'Release C', which consistently exhibit high sojourn times and influence patient outcomes significantly. These activities emerge as bottlenecks in the patient care process, particularly in cases of severe sepsis, where delays can lead to increased complications and mortality. By identifying these critical points, the study provides actionable insights for healthcare providers to optimize resource allocation, reduce delays and enhance the overall efficiency of sepsis management. The findings underscore the importance of targeted interventions in these key areas, offering a path toward improved clinical outcomes and reduced sepsis-related mortality and readmission rates. This research contributes to the growing field of process mining in healthcare, highlighting its potential to transform complex clinical pathways into more efficient and effective treatment processes.