9th European Medical and Biological Engineering Conference

Computer-aided detection (CADe) for endoscopy can help physicians to locate and identify lesions better, but there are still many false positives (FP) when processing cross-domain data. This paper proposes SE-SORT, a plug-and-play tracker, designed to be seamlessly integrated as a post-processing plugin into existing CADe systems to enhance the adaptability to cross-domain data. The proposed tracker adds trajectory initialization thresholds into the tracking association strategy, reducing the impact of high confidence FPs on the matching process. Experiments show that the modified tracker effectively reduces the impact without significantly affecting the processing speed. This allows the detector of CADe system to tolerate lower detection confidence thresholds, thus improving the overall accuracy on cross-domain data. Compared to existing SORT trackers, the proposed tracker exhibits better accuracy and higher efficiency in endoscopic lesion detection and tracking. This work will help to improve the generalization and expanding the clinical application scope of related works on endoscopic real-time CADe.

This article discusses the use of a combination of mutation status of the immunoglobulin heavy chain (IGHV) gene and entropies of DNA sequence for the analysis of leukemia patients’ survival. Using a chronic lymphocytic leukemia (CLL) patient database, the study calculates different entropy measures for each patient’s DNA sequence. An entropy maximization algorithm is developed to estimate the statistical DNA information of each patient, allowing for the classification of patients into two groups without relying on population properties. Survival analysis of leukemia patients is conducted by combining IGHV subtype analysis and entropy measurements. Statistical significance is found when comparing groups with high and low entropy, as well as different IGHV subtypes. The analysis indicates that the combination of a mutated IGHV subtype and high entropy of DNA sequence can significantly impact the life expectancy of leukemia patients. The results are validated using Kaplan-Meier survival analysis, Cox regressions, and a Generalized Linear Mixed Model (GLMM). Overall, the study demonstrates the value of combining IGHV gene mutation status and entropy analysis of DNA sequences for leukemia patient survival prediction. The findings highlight the potential for leveraging multiple prognostic factors to improve the identification of patients with varying lifespan prognoses.

Animal models undeniably offer advantages for studying bone regeneration in bone tissue engineering. Currently, a lack of documented and standardized critical size defects (CSD) protocol exists for femoral bone. This study established a femoral bone rat model to evaluate engineered scaffold and its effect on bone regeneration. Eight Sprague-Dawley rats were divided into four groups, each induced with specific sizes of circular femoral defects: 1.5 mm diameter; 4.0 mm depth (Groups A and B), and 2.4 mm diameter; 7.0 mm depth (Groups C and D). Rats were euthanized at 4- and 8-weeks post-induction. Observations revealed that the 4-week period was insufficient for initiating the bone healing process. Notable signs of bone healing and remodelling become apparent only after 8 weeks with normal morbidity scoring at week 5 onwards. Gross examination indicated that rat models with a defect size of 1.5 mm diameter; and 4.0 mm depth healed at a faster rate suggesting inadequate defect size. In contrast, the rat model 2.4 mm diameter; and 7.0 mm depth defect emerged as the suitable model with evidence of newly formed bone signifying the process of mineralization at the defect site. The Hematoxylin and Eosin (H&E) staining of bone tissue demonstrates substantial formation of bone tissues (osteoid) and vascularized areas, consequently supporting the efficacy of this model. Therefore, this study finds that the 8-week timepoint with a 2.4 mm diameter and 7.0 mm circular defect is ideal for assessing bone regeneration of an engineered scaffold in rat bone model.

Tympanometry is a gold standard method for the evaluation of hearing function, allowing the identification of pathological alterations of the outer and middle ear by a non-invasive approach. However, in order to make the measurement it is necessary to alter the pressure of the outer ear, which limits the use of the technique for newborns and people with tympanic perforation. To overcome this problem, a complete pressure less technique (PLAI™) was proposed. This paper aims to present the statistically significant correlations between the resonance frequency measured with PLAI™ and the estimated volume and compliance obtained through tympanometry in both healthy subjects and patients affected by Otitis Media with Effusion (OME). The tests, conducted on 57 adult subjects, indicated a significant linear relation between the volume measured with tympanometry and the resonance frequency obtained with PLAI™ that could be used to calculate a value of volume comparable with the gold standard. Moreover, an inversely proportional relation between the compliance from tympanometry and the resonance frequency from PLAI™ was found, even if affected by several outliers, which hinder a practical use of this specific relation. These two preliminary findings show that it is possible to use this new technique as a preliminary test in subjects deemed to be at risk with tympanometry, eliminating the drawbacks related to the pressure changes but offering a comparable measurement of volume and possible future chances of research.

The objective of this study was to explore whether cross-entropy metrics provide further insights in comparison to conventional entropy, aiming to enhance the accuracy of mental workload assessment. We initially filtered and segmented EEG signals from two prefrontal channels and decomposed the signals into subbands. Afterward, we calculated a range of cross-entropy and traditional entropy metrics for each sub-band. Finally, these extracted features were fed into an AdaBoost classifier to evaluate mental workload levels. The comparison of classification results demonstrated that integrating cross-sample entropy with sample entropy notably enhanced accuracy by 10%, reaching 84%. Further, utilizing the complete set of cross-entropy metrics yielded an accuracy of 84.5%.

There is significant potential in the treatment of osteoporosis using hydrogels tissue scaffolds impregnated with cells, growth factors and antibiotics. However, the delivery of hydrogel into the bone is yet to be developed. Non-invasive delivery is desirable for both patients and healthcare systems. Aim: evaluate hydrogel flow mechanisms and how that can affect the bone structure as well as the impregnated cells to begin the understanding of injection delivery. Methods: hydrogel flow dynamics are investigated using a linear laminar flow computational fluid dynamics model. Three cases with different Reynold’s (Re) numbers were simulated and trabecular bone wall shear stress, and hydrogel flow velocity were extracted to evaluate bone/cell damage and velocity profile. Results: higher Re number which is proportional to higher inlet velocity is associated with higher bone wall shear stress, and hydrogel pressure. Although the shear stress and pressure do not pose any risk regarding the bone structure, the cell viability is compromised in the case of Re = 6. Conclusions: the outcome of this study provides insights that will contribute to the design of a hydrogel injection procedure. Further work is needed to improve the model to further inform hydrogel delivery mechanisms.

The European Health Data Space (EHDS) introduction has raised a discussion on the use of health data (HD) for secondary purposes that could essentially improve health systems efficiency. This research examines how key stakeholders in the Czech Republic feel about this process. In qualitative analysis (interviews), two major themes were identified concerning the use of health data: Risks and Barriers, as well as Health Data Access theme. We investigated the views and perspectives of 8 key stakeholders from healthcare sector (insurance company representatives, Ministry of Health, IT technologies, data applicants etc.). These opinions were analyzed using the software MAXQDA 24 and compared to better understand the challenges and benefits associated with using health data for secondary purposes. According to our study, the benefits surpass the risks and barriers shown by the secondary utilization of health data. The stakeholders highly appreciated making this data accessible to enhance quality care and efficiencies in the healthcare system. Therefore, this study’s outcomes are an aid in understanding and promoting of use of health data within EHDS confines. Furthermore, they will be valuable not just to researchers but also to policymakers, businesses, and the public interested in harnessing health data to enhance the healthcare system.

This paper presents cluster analysis of tri-axial accelerometer data acquired from various human physical activities as well as simulated falls using frequency domain features. Clustering was performed using K Means, Gaussian Mixed Model and Fuzzy C-Means methods. In our analysis we focused on two problems: the first clustering problem being activity recognition and differentiation from simulated human falls, while the other problem focused on distinction between single jerk events (e.g. jumping, falling) and continuous activity signals (e.g. running, walking).

This study aims to implement a pattern classification algorithm for plantar thermograms, focusing on identifying altered temperature zones in the feet of diabetic patients. Utilizing a database of 334 thermograms, various classification algorithms including Support Vector Machine (SVM), Logistic Regression, Artificial Neural Network (ANN), Random Forest, and K Nearest Neighbors (KNN) were evaluated. Features extracted from the literature, such as number of pixels, maximum entropy, variance, mean value, correlation, contrast, energy, and homogeneity, were utilized for training and evaluation. The classification task involved assigning thermograms to 5 classes based on the thermal change index (TCI).Performance evaluation was conducted using an information theory metric approach based on mutual information, measuring the alignment between predicted and true classes. The neural network achieved the highest mutual information score of 2.69 out of 5, indicating that approximately 53.8% of the information obtained from model predictions aligned with the true classes. Additionally, a database was established in the Footlab BASPI laboratory, comprising 20 thermograms from the plantar region of 10 control subjects. A novel protocol, incorporating additional elements to the STANDUP base protocol, was proposed.Finally, classification using the ANN on data acquired from the Footlab - BASPI database yielded satisfactory results, successfully distinguishing between 1.7 classes, representing an 85% success rate in classifying thermograms.

The shoulder joint is susceptible to be damaged in sports with overhead actions, often leading to swimmer shoulder pathology. Fatigue can also worsen and increase the risk of overuse injuries. Evaluating shoulder kinematics during swimming is crucial to identify injury-related movement patterns and to provide a correct physiotherapy treatment. To measure kinematics, inertial and magnetic measurement systems (IMMSs) offer a very versatile approach with respect to traditional video-based systems. This preliminary study focuses on the effects of fatigue on shoulder joint kinematics in swimmer with swimmer shoulder compared to healthy swimmers, by using IMMS. 11 young swimmers (5 pathological, 7 male) took part in the study. Each participant executed 40 s of dry front crawl followed by a fatiguing protocol and by other 40 s of dry front crawl. We analyze the arm movement relative to the thorax examining the differences of the movement amplitudes between healthy and pathological subjects and before and after fatigue exercise according to the three rotations: Flexion/Extension, Abduction/Adduction, and Internal/External rotation. Some slight non-significant differences were found after exercise compared to before in all the three rotations while a significant difference between healthy and pathological subjects was found in Flexion/Extension rotation both before and after fatigue exercise. The use of IMMS allowed to verify the repeatability of the kinematic movement and to quantify the rotation angles identifying which component of the movement is most affected by the swimmer shoulder pathology. However, a larger number of subjects is necessary in order to confirm the results.

This study explores the Gaussian Copula Synthesizer’s (GCS) utility in addressing the limitations of a small dataset (58 real patient records) in Atrial Fibrillation (AF) research, focusing on Heart Rate Variability (HRV). Leveraging this method, we generated a realistic synthetic dataset of 1, 000 records, replicating the features observed in the original records. The GCS effectively expands dataset size while maintaining HRV pattern realism. This aids in developing and refining models used in AF research, overcoming challenges associated with limited sample sizes. Emphasizing privacy considerations, this approach showcases the potential of classic statistical methods in synthetic data generation for advancing AF research within the constraints of small datasets.

The dynamic field of transitional care is witnessing a transformation, fueled by the integration of innovative technologies and methodologies designed to enhance patient outcomes and streamline healthcare processes. Among these, the role of Living Labs as open innovation environments is becoming increasingly significant, offering a novel infrastructure for creating harmonized methodologies for the collection of data from Internet of Things (IoT) health devices. This paper delves into the role of Living Labs in shaping the future of data collection in transitional care, focusing on the synergy between user-centered innovation ecosystems and the seamless integration of diverse technological solutions. This paper presents a comprehensive analysis of the methodologies and strategies implemented by Living Labs to facilitate the harmonized collection of IoT health device data. Through this study, we explore the potential of Living Labs to drive innovation in transitional care by providing a controlled yet flexible environment for experimenting with new technologies and data collection techniques.

In the current study, an analysis of mobile health (m-health) technology usage in managing prostate cancer, focusing on patients undergoing radiotherapy, was conducted. Amid increasing incidence and mortality rates of prostate cancer in the Czech Republic, exploring m-health potential in oncological care is of critical research interest. The aim was to identify the benefits of m-health in radiotherapy, assessing current market dynamics and stakeholder roles to inform the conceptualization of future m-health application development in oncology. A qualitative approach utilizing semi-structured interviews (using software MAXQDATA 22) with patients and medical staff was employed to gain deeper insights into personal experiences and preferences regarding m-health applications and devices. The analysis showed limited patient engagement with m-health technologies; some were interested in health monitoring, while others were worried about complexity and usability. Medical professionals recognized m-health’s potential in treatment optimization but noted integration and data security challenges. In conclusion, m-health represents significant potential for innovation in prostate cancer patient care. However, addressing existing challenges, including ensuring user-friendliness, effective integration into healthcare systems, and protecting personal health data, is imperative. These findings provide an overview of the current state and outline potential trajectories for future development and implementation of m-health in oncology.

Spatial analysis of EEG data, e.g. using short-term stable microstates, is increasingly used in neuroscience and clinical applications. At the same time, scenarios involving mobility, sports, and home-based activities are becoming more prevalent in EEG studies. For this purpose, dry EEG electrodes are more and more commonly used. Thus, our objective was a comparison of microstates analysis between dry and gel-based EEGs. 256-channel EEGs were recorded from 30 volunteers using dry and gel-based electrodes during resting state eyes closed and eyes open. Microstates were extracted for each measurement and time-domain parameters were calculated. We found a high degree of consistency between the microstate maps extracted from dry and gel-based measurements for both eyes closed and eyes open conditions. The topographic similarities between the average maps for dry and gel-based recordings were above 81.5% for each of the seven extracted maps. We conclude that topographic microstate analyses are feasible using multichannel EEG setups with new dry EEG electrodes.

Photoplethysmography (PPG) is an optical technique used for the continuous monitoring of blood volume changes, and is integrated in a variety of medical devices and consumer wearables. Meanwhile, in silico and in vitro experiments have shown to facilitate our understanding of light-tissue interactions to assess the feasibility and accuracy of PPG-based technologies in acquiring PPG signals from the human anatomy. Therefore, this study demonstrates an approach to validate an in vitro study of a developed wrist phantom through a Monte Carlo (MC) simulation to compare PPG signals generated from experimental and computational models. The MC model simulates a PPG signal using MATLAB, and accounts for the optical and mechanical properties of the phantom and vessel in synchrony with changes in internal blood pressure. Key features of the PPG signal, such as onsets, and systolic and diastolic slopes are visually apparent. Additionally, the use of in-built MATLAB functions are recommended to improve the morphology of the PPG signal to optimise computational resources. Overall, the current study shows the applicability of Monte Carlo simulations to validate PPG data acquired from in vitro models.

The brain-computer interface (BCI) is a communication method between the brain and computer or electronic devices. An electroencephalography (EEG) signal is a popular way to create a traditional BCI system. However, EEG signals can be sensitive and prone to interference from both external and internal noise, complicating system design and potentially leading to malfunctions in the BCI system. Near-infrared spectroscopy (NIRS), which is less susceptible to external noise, complements EEG in BCI systems. However, the sampling frequencies of EEG and NIRS signals differ significantly. It needs some complicated process to combine them or design a system that considers the characteristics of both data. This study explores the use of a flexible convolutional neural network (CNN) architecture to mitigate the impact of disparate sampling frequencies on accurately classifying left- and right-hand movements using raw EEG and NIRS data. By employing a multi-input CNN approach, the study aims for significant improvements in classification accuracy. The results suggest that the proposed model achieves high accuracy in classifying tasks using participant-dependent cross-validation. Therefore, the proposed model can classify a label with high accuracy, and the utilization of multi-input methods supports addressing measurements-related issues, such as temporal resolution, in BCI systems. EEG is superior to NIRS in estimating the localization of brain function. However, NIRS excels in estimating relative changes in blood flow. Creating a system that exploits both advantages is essential for BCI applications.

Non-invasive fetal ECG extraction from abdominal signals might provide significant information for long-term fetal monitoring, being very attractive for physicist. Nevertheless, accurate extraction of the fetal ECG is a challenging task, due to the disturbing signals, which overlap the signal of interest in the frequency domain. Among the current denoising methods, neural networks are very attractive due to their performance. The current paper proposes a linear feed-forward neural network that estimates very accurately the abdominal mECG, the strongest disturbing signal, based on two thoracic mECG, removing it thereafter. The obtained results are very promising, allowing the further investigation of the fHR, for the fetal well-being evaluation. The comparison with the event-synchronous interference canceller shows the advantage of the neural network in preserving the fECG morphology, with the cost of higher computation time. Both methods require the preprocessing of abdominal signal in order to remove the power line interference and the baseline wander.

Abdominal compartment syndrome (ACS) is characterized by progressive intra-abdominal organ dysfunction resulting from elevated intra-abdominal pressure (IAP). Measuring the IAP with periodic intervals is essential for timely intervention, i.e., for keeping the IAP at normal levels. The current clinical method is invasive and offers only intermittent assessments, limiting its effectiveness in continuously monitoring the IAP. Here, we present a continuous multimodal IAP monitoring system (IAP-CMM), which is non-invasive and provides continuous readings. The IAP-CMM device is equipped with bioimpedance measurement (BioZ) and a mechanical muscle contraction force (MC) sensor. Electrical measurements are performed via dry electrodes. Following performance and safety verifications, the device was tested on 4 patients who underwent laparoscopic surgery. The preliminary results suggest MC has a high linear relationship with IAP (r = 0.99; r2 = 0.987) and the BioZ exhibits a second order relationship with IAP (r2 = 0.95). An extensive clinical study recruiting more patients is needed to draw statistically meaningful conclusions.

The paper outlines the OHIO project, funded by the European Union’s Horizon 2020 research and innovation action programme, via the ODIN - Open Call issued and executed under the ODIN project (GA 101017331), which focuses on the enhancement of hospital safety, productivity, and quality. The main objective of OHIO is to provide a solution to empower the management of hospital facilities in terms of clinical engineering and logistics for the use-case hospital “Le Scotte” in Siena, Italy. The OHIO framework integrates various technical solutions to address possible management challenges which may arise during the maintenance of medical equipment, such as the unavailability and the untraceability of a device, or the technicians’ lack of knowledge on how to reach it. The paper also describes a set of Key Performance Indicators (KPIs) designed for measuring OHIO’s impact on maintenance timings and efficiency. Preliminary outcomes show promising results in preventing failures, improving scheduling, and providing efficient indoor navigation. The OHIO project demonstrates potential enhancements in hospital operations and maintenance through digital IoT solutions, successfully responding to a set of previously unmet needs.

This study aimed to estimate concentration of uric acid (UA) in blood from online spent dialysate measurements during haemodialysis (HD) with an optical sensor non-invasively. Twenty-two patients on chronic (HD) were monitored during 88 treatment sessions, including HD and haemodiafiltration treatments. Pre- and post-dialysis blood samples and spent dialysate samples, from the drain outlet of HD machine, were collected during each session. HPLC analysis was used as a reference to determine UA concentrations in the samples. An optical sensor was connected to the drain outlet of the HD machine and ultraviolet light absorption of spent dialysate was measured online for each HD session at four different wavelengths. A linear interactions regression model was used to estimate UA concentrations in spent dialysate with the optical sensor based on the light absorption. Sensor estimated spent dialysate UA concentrations were highly accurate with standard error (SE) < 3.50 µmol/L and strongly correlated (R2 >= 0.983) to the actual UA concentrations of spent dialysate in the training and the test dataset. A linear regression model, using treatment settings and UA concentration determined with optical sensor, was employed to estimate concentration of UA in blood samples. Sensor estimated blood UA concentrations were similarly highly accurate (SE < 32.1 µmol/L) and strongly correlated (R2 > 0.95) to the actual UA concentrations of blood in the calibration and the validation dataset. In conclusion, this study showed that UA levels in blood can be estimated noninvasively with the optical sensor in real time, with higher accuracy than shown previously.

Use and understanding of electroporation have grown in recent years, revolutionizing various fields. However, optimization for stimulation techniques is still needed. In this context, the introduction of genetically modified cell cultures allows a dramatic increase in simulation capabilities, but also introduces the necessity of more advanced and human independent analysis methods.We aimed to identify features, including morphological characteristics and other experimental parameters, to develop models for predicting the responses of genetically engineered HEK (Human Embryonic Kidney) cells to pulsed electric fields.This subset of predictive features, including the presence of K+ channels, electric field strength, experiment number, initial fluorescence, and cell morphology characteristics was identified.A machine learning approach based on ensemble learning techniques deployed through the XGBoost algorithm was utilized. This approach involves sequentially building numerous weak decision trees, where each subsequent tree aims to correct the errors made by the ones before it. Considering the unbalanced frequencies of the cell response types, we adopted different strategies to balance the training set and avoid bias were adopted.The produced XGBoost model trained with a combination of real and synthetic data exhibited an accuracy of 66.0%, a mean AUC of 0.89, and an average F1 score of 0.66 when evaluated against the internal test set comprising solely real data. Further analysis on an external test set revealed an F1 score of 0.57.In conclusion, we identified predictive features and produced models that may contribute to predicting the responses of genetically engineered HEK cells to pulsed electric fields.

Hepatitis C Virus (HCV) infection is a significant global health concern with approximately 1.5 million new infections yearly. The choice of the most appropriate HCV treatment depends on several factors, including liver fibrosis status. Current guidelines recommend liver fibrosis evaluation using non-invasive techniques such as Liver Stiffness Measurement (LSM) using liver elastography. Although LSM revolutionized patient care in the last decade, allowing biopsy-free treatments, several factors can lead to overestimation or underestimation of liver stiffness values, affecting management strategies. This study presents a machine-learning approach using an eXtreme Gradient Boosting model to predict possible LSM inaccuracies in a cohort of 509 HCV-positive treated patients. The dataset, characterized by 55 variables, underwent feature reduction and balancing to mitigate class imbalance to train the predictive algorithm. The developed model can identify inaccuracy in LSM and achieves an accuracy of 88.0% on the training set and 92.0% on the test set. Furthermore, it exhibited a consistent mean Area Under the Curve One-vs-One (AUC-ovo) of 0.97 across both datasets. The model’s performance in predicting abnormal LSM may enable healthcare providers to tailor treatment plans more precisely, optimize patient follow-up, and reduce unnecessary invasive procedures. These findings highlight the potential of machine learning in improving patient care in the context of chronic HCV management.

Machine learning has made significant advances in many areas, particularly in the healthcare domain. However, despite the advances, the implementation of these models in clinical scenarios is still limited due to several challenges, including the lack of trust. Standard performance measures, such as sensitivity, specificity and confidence intervals can be used to evaluate the reliability of a model, but these are overall performance metrics and do not provide insight into the performance of individual instances. Moreover, these estimates are typically calculated during the training phase and are not easily generalized to new, unseen instances, occurring in the deployment phase.As result, besides the prediction outcome, the existence of a measure of reliability in the prediction of individual estimations would add a layer of security, increasing trust in human-AI interaction, as well as it could also be helpful to support the improvement of the model. This study proposes a reliability measure, combining density and local fit principles, to estimate the confidence of individual predictions in the deployment phase. When applied to a machine learning model in the cardiovascular risk assessment context, the method demonstrates the ability to distinguish between reliable and unreliable predictions, as well as aiding in the stratification of the population.

Depression is a serious medical illness that adversely affects how a person feels, thinks, and behaves. This illness can be treated with the aid of Transcranial Direct Current Stimulation (tDCS), which can help to reduce the symptoms of depression. The level of illness is typically evaluated using the Hamilton Depression Rating Scale (HDRS). The focus of this paper is the prediction of the HDRS score after a tDCS course. By predicting the result of tDCS, psychiatrists can provide better counseling to the patients about their future conditions after the treatment and decide wisely about the treatment method. We used different kinds of demographic information, treatment information, and the HDRS score before the treatment as predictors and supervised Machine Learning (ML) algorithms for the prediction. The analysis is conducted on 169 patients with depression. Our preliminary results show that the accuracy can be up to 63% when predicting the value of HDRS after tDCS treatment sessions as a binary variable using Gradient Boosting. This is encouraging on such a small data set. Moreover, our results provide insight into the predictors pivotal to this outcome. They show that the HDRS score at baseline, the age, and the gender of the subject are the three main predictors. The results suggest this methodology may yield very interesting results.

The veracity and reliability of information in the health sector is essential for security in the Predictive Health Technology Management (HTM) and quality in patient care in health care establishments. One technology that has been contributing to the traceability and security of data is Blockchain, which consists of a decentralized network for storing transactions in an immutable way. The objective of this work was to develop a study of the intersection between blockchain technology and the application of HTM by Clinical Engineering. A rapid literature review of blockchain applications in medical devices was developed and presented two models of integration of this technology in hospitals. The first proposal was about the incorporation of Blockchain in the traceability of information incorporated in the historical record in the lifecycle of the technology and the second in the application in calibration certificates and test reports on medical equipment.

Hospital environments are becoming more and more complex due to technological advancements. The allocation of resources, especially during events like the SARS-CoV-2 pandemic, is becoming crucial, leading Clinical Engineering and Health Technology Management professionals to adopt evidence-based strategies for maintaining the reliability of medical equipment. Real-World Data (RWD) is leveraged to generate Real-World Evidence for assessing the effectiveness and safety of health technologies. Evidence-based maintenance, a systematic process crucial for superior healthcare services, uses empirical RWD to identify effective maintenance strategies. The study aims to propose a standardized XML schema for data exchange, linking maintenance activities with hospital systems while incorporating maintenance data classification. This approach aims to overcome data-sharing challenges while providing insights into maintenance effectiveness and assessment.

Two rare genetic eye disorders, known as Retinitis Pigmentosa (RP) and Stargardt Disease (STGD), both falling under the classification of Inherited Retinal Diseases (IRDs), have emerged as focal points of investigation in the pursuit of potential treatments leveraging cutting-edge technologies, notably artificial intelligence (AI) integrated with fundoscopy. These IRDs, characterized by their genetic underpinnings, serve as poignant reminders of the intricate and multifaceted nature of genetic eye disorders. The primary objective of this work was to develope an algorithm capable of automatically categorizing fundoscopies obtained from 74 pediatric eyes. In pursuit of this goal, an artificial intelligence algorithm was developed, exploiting the YOLOv8n Net. Through rigorous testing, it was demonstrated that this algorithm effectively and accurately classified the samples within the test set, exhibiting a notable absence of misclassification errors. The overarching ambition of this study is to introduce a robust and reliable classification tool that can significantly enhance the diagnostic process for rare diseases such as RP and STGD. By leveraging the power of advanced technologies, particularly AI, this research endeavors to streamline and optimize diagnostic procedures, thereby offering hope for improved management and treatment outcomes for individuals affected by these challenging conditions.

As AI technologies progress rapidly, there is an increasing need for tailored regulations that effectively address data provision, sharing, utilization, and knowledge generation. This paper delves into the essential regulations and emphasizes the crucial role of AI model validation in guaranteeing the dependability and effectiveness of AI-driven solutions. An innovative approach is introduced, detailing an organized four-phase methodology for external validation. The integration of these frameworks and the implementation of a DataLab are deemed imperative for fostering transparency, accountability, and enhancing patient outcomes within the swiftly evolving landscape of AI in healthcare. Through a comprehensive examination of key regulations and a structured validation approach, this research underscores the critical need for meticulous scrutiny and validation of AI models to ensure their reliability and efficacy in improving healthcare delivery. This study aims to lay the foundation for further exploration and advancement in this pivotal area, offering a roadmap for stakeholders, researchers, and policymakers to navigate the complexities of AI integration in healthcare while prioritizing patient safety and quality of care. The work has been done in the framework of the GATEKEEPER project, funded by the European Commission under the Horizon 2020 program.

The correlation between electroencephalographic (EEG) and electrocardiographic (ECG) signals can provide important information about cardiovascular regulation. The current study aims to investigate the dependence of the correlation between EEG and heart rate variability (HRV) on cardiovascular indicators. The signals of a group of 30 subjects were divided into two groups of 15 subjects according to the lower and higher indicators of blood pressure and cholesterol. Relative EEG frequency band powers were calculated in theta, alpha, and beta frequency bands. From power spectral analysis of HRV, low frequency (LF) power, high frequency (HF) power and LF/HF were calculated. In the current study, the correlation between EEG and HRV is detected by two HRV features, LF and HF, in the theta band, and also by two features, HF and LF/HF, in the alpha band. The correlation by one HRV feature, LF/HF, is revealed in the beta band. The effect of cardiovascular features is reflected by LF and HF features in the theta band and by HF in the alpha band. The novel finding that even a small increase in cardiovascular features (blood pressure and cholesterol) can affect cardio-neuronal regulation is important and needs further investigation.

The main goal of the study is to investigate materials from dental practice and assess their suitability for manufacturing anthropomorphic phantoms designed for X-ray imaging modalities. For this purpose, we investigated nine commercial materials commonly employed in dental procedures for tasks such as restoration, duplication, and imitation. They underwent scanning at a clinical Computed Tomography (CT) facility using six anode voltages. Subsequently, we measured their CT numbers, expressed in Hounsfield Units (HU). The findings of our investigation indicate that three specific materials, namely Temp silic, Gingifast Elastic, Ellite Double A, exhibit promising characteristics for the production of anatomically accurate bone structures in anthropomorphic phantoms.

Tele-rehabilitation has the potential to transform the way patients are monitored from home, overcoming geographical barriers, enhancing accessibility, and promoting patient autonomy. The development of a tele-rehabilitation system capable of automating the recognition of performed exercises may significantly impact rehabilitation outcomes. Implementation of machine learning algorithms combined with magneto-inertial measurement units (M-IMUs) has enabled remote home-based rehabilitation therapy through wearable systems. Thus, in this study sixteen healthy participants and sixteen patients with rotator cuff injuries were enrolled to perform six shoulder rehabilitation exercises while wearing a wearable system based on three M-IMUs. This study aimed to conduct a thorough analysis of the features extracted from time-series data collected by these three sensors during these exercises. The statistical analysis indicated statistically significant differences in task features, but not between participant groups. Three features, identified as the most representative and distinctive among all tasks, were subsequently, used to train the Support Vector Classifier in classifying the six exercises. The obtained classification results are promising for the application of this wearable device in remote monitoring of patients with shoulder musculoskeletal disorders during home-based rehabilitation exercises. Further studies will involve the implementation of the Principal Component Analysis (PCA), along with the training of additional machine learning models.

Measurements of the oxygen content in fluids and gases (oximetry) have found many applications in various fields, and are particularly desirable in medicine and environmental protection. The standard method for measuring of oxygen tension, also known as oxygen pressure (pO2) in fluids and gases, including blood and respiratory gases, is the Clark's electrochemical method. The authors have developed and tested a simple electrochemical sensor based on the Clark method for measuring oxygen pressure in blood and respiratory gases. A sensor, developed by us for the assessment of pO2 by the amperometric method, meets the basic metrological requirements for use in direct oxygen tension measurements of blood and respiratory gases.

Texture analysis of breast ultrasound images is widely used in the tissue characterization of breast tumors, as the texture features are closely connected to the scattering patterns and microstructures of the tumors. H-scan imaging is a novel ultrasound imaging method to differentiate the backscattering patterns inside different tissue structures. In this work, the feasibility of H-scan image texture analysis for the characterization of breast tumors is assessed. Five texture features (Contrast, Energy, Entropy, Homogeneity, and Correlation) are obtained from the H-scan images based on the Open Access Series of Breast Ultrasonic Data (OASBUD) involving 100 breast tumors of 78 female patients. Texture analysis results demonstrated significant differences (p < 0.05) between benign and malignant breast tumors for the selected features derived from H-scan images. Moreover, several textural features (Energy, Entropy, Correlation) from H-scan images began to show significant differences between certain BI-RADS category levels (level 4a with level 5, level 4b with level 5). In contrast, no significant difference was in the features extracted from B-scan images of the aforementioned BI-RADS category groups. Overall, this study demonstrates that texture analysis of H-scan images is helpful for characterizing breast tumors.

This study introduces a framework for establishing a living lab for the aggregation of data, scenarios, and ecosystems to improve adherence to medical treatments. Recognizing the complexity of adherence behaviours across different health conditions, the living lab environment enables 1) the management of real-world data, 2) the definition and simulation of real-world scenarios, and 3) the implementation of real-world solutions where an intervention to address adherence is introduced. The framework may have potential in bridging the gap between theoretical research and real practice, offering a scalable solution to improve health outcomes by optimizing adherence to treatment.

Sleep quality is a vital component of one’s overall health and well-being. Inadequate sleep quality is linked to various adverse consequences, including cognitive decline, mood disruptions, and an elevated susceptibility to non-communicable diseases. Hence, it is crucial to precisely evaluate the quality of sleep, in order to identify individuals who are at risk and to develop successful interventions. Importantly, it has been shown that sleep quality can impact physiological processes even when a person is awake, leading to changes in heart rate variability (HRV). From this standpoint, the utilization of wearables and contactless technologies that can measure HRV without causing any discomfort is extremely well-suited for evaluating sleep quality. Nevertheless, there is a dearth of studies that analyze the correlation between HRV and sleep quality during waking. The aim of this study is to create a machine-(ML) learning model that uses HRV data to estimate sleep quality, as evaluated by the Pittsburgh Sleep Quality Index (PSQI). The measurement of HRV was conducted using a wearable photoplethysmography (PPG) sensor positioned on the fingertip. Subsequently, models were created to classify sleep quality based on the PSQI score. By employing the current approach, a classification good accuracy of 76.7% was achieved. In summary, this study has the potential to facilitate the use of wearable and contactless technology for monitoring sleep quality in ergonomic applications.

Deep vein thrombosis (DVT) is the formation of a blood clot within the deep veins, most commonly those of the lower limbs, causing obstruction of blood flow. In 50% of people with DVT, the clot eventually breaks off and travels to the lung to cause pulmonary embolism. Clinical assessment of DVT is notoriously unreliable because up to 2/3 of DVT episodes are clinically silent and patients are symptom free even when pulmonary embolism has developed. Early diagnosis of DVT is crucial, and despite the progress made in ultrasound imaging and plethysmography techniques, there is a need for new methods to enable continuous monitoring of DVT at the point of care. This paper presents the conceptual design and methodology towards a novel wearable diagnostic device for point-of-care, operator-free, continuous monitoring in patients with high DVT risk. The device will combine novel wearable hardware for ultrasound imaging and impedance plethysmography with autonomous, AI driven DVT detection, to allow continuous monitoring for blood clot formation in the lower limb. Activity and other physiological measurements will be used to provide a continuous assessment of DVT risk and guide the automated scanning via an intelligent decision support unit that will provide accurate monitoring and alerts. The work is supported by the Horizon project ThrombUS+ co-funded by the European Union. (Grant Agreement No. 101137227).

Brain-Computer Interfaces (BCIs) offer direct communication between the brain and external devices, holding immense potential across various applications. This study focuses on Motor Imagery-based BCIs (MI-BCI), decoding neural patterns associated with mentally rehearsed motor actions. Despite their promise, BCIs face challenges in real-world applications, primarily in reliability and complexity. While classification accuracy is a standard metric for BCI performance, the literature often overlooks real-time responsiveness. Many studies report classification outcomes offline, disregarding the prompt translation of EEG signals into actions. The acceptable delay from EEG signal to action should not exceed 1 s; however, numerous studies employ time-windows exceeding 4 s, affecting user control perception. This article aims to compare the trade-off between time-window length and classification accuracy in MI-BCI, using three linear classifiers (LDA, MLP, SVM). Participants include stroke patients and subjects from the BCI IVa dataset. Results demonstrate time-frequency plots indicating MI-related EEG changes, revealing a trade-off between accuracy and responsiveness. Our findings underscores the importance of addressing real-time responsiveness in BCI evaluations, proposing a balance for practical system utility. In conclusion, this study enhances our understanding of the delicate balance needed for optimal real-world application of MI-BCIs, emphasizing the trade-off between accuracy and responsiveness.

Age-related macular degeneration (AMD) is the leading cause of vision loss in the elderly population. Transfer learning has proven useful in fundus image analysis for early diagnosis. Previous models were pre-trained on the ImageNet database. However, a source domain related to retinal diagnosis would facilitate model learning. Our objective was to apply transfer learning from the domain of diabetic retinopathy (DR) to aid in the detection of AMD (binary classification). The proposed model was based on the ResNet-RS architecture. Pre-training aimed at DR diagnosis was conducted using the Kaggle database. Then, fine-tuning was performed using the Automatic Detection challenge on Age-related Macular degeneration (ADAM) dataset. We carried out 3 experiments with different number of images used for fine-tuning. As the main result, our method showed a much faster convergence than the corresponding models pre-trained on ImageNet. Additionally, the proposed source domain was proven especially useful when scarce data in the destination domain was available.

This paper examines the regulatory challenges encountered by Tremitas GmbH, an Austrian startup that developed the Tremipen, a device for quantifying tremor, within the European Union’s medical device sector. Transitioning from the Medical Device Directive (MDD) to the Medical Device Regulation (MDR) introduced complexities that disproportionately affect startups due to stringent compliance requirements. Through a detailed postmortem analysis, we identified critical regulatory missteps by Tremitas, including inadequate understanding of the regulatory framework, mismanagement of clinical trials, and ineffective outsourcing of regulatory activities. These challenges culminated in delayed market entry and contributed to the company’s eventual insolvency. Key results indicate that early integration of regulatory knowledge, strategic planning for clinical validation, and careful selection of development partners are essential for navigating the medtech regulatory landscape. The paper underscores the importance of regulatory affairs in the success of medical device startups and suggests that a proactive approach to compliance can mitigate risks and enhance market success. This analysis aims to serve as a cautionary tale and guide for future startups in the medical device industry, emphasizing the critical role of regulatory strategy in entrepreneurial success.

The significance of emotional assessment has gained increasing recognition across diverse fields, such as psychology, healthcare, education, and social sciences. It is deemed crucial for comprehending and addressing a broad spectrum of outcomes, including mental health, academic performance, patient experiences, and social acceptance. A noteworthy aspect of human emotions lies in their expression through a diverse range of vocal sounds, which can be interpreted and understood by the listener. From this perspective, the application of Mel Frequency Cepstral Coefficients (MFCCs) and Gammatone Frequency Cepstral Coefficients (GTCCs) emerges as a brilliant approach to portraying the concealed vocal components that express emotional states. In this study, we employed Machine Learning (ML) techniques to leverage MFCCs and GTCCs, aiming to construct a classifier capable of assessing different emotions. We utilized the freely available Toronto Emotional Speech Set (TESS), a dataset comprising vocal recordings of two actresses repeating 200 words in seven distinct emotions. Furthermore, we introduced an age-related analysis to gain insights into how age might impact the human capacity to express emotions through vocal expression. Results obtained showed an accuracy of 99.6% in the emotional assessment for vocal recordings of both actresses employing GTCCs. Although the study is focused on the investigation of only two individuals, it represents an initial step in understanding the potential influence of age on emotional expression through vocalization. Furthermore, the age-related studies presented a 100% accuracy in the emotional assessment restricted to the vocal samples from the younger actress, compared to the 98.6% obtained in the one restricted to the older actress. These findings suggest that the age could influence the emotional expression.

An estrogen deficiency at menopause induces climacteric symptoms in humans including hot flashes, diaphoresis, sleep disturbances and progressive bone loss. A phytoestrogen can be defined as a plant-derived xenoestrogen molecule with estrogenic activity and not associated with the endocrine system. The present study aims to investigate the chemical constituents and their potency as a phytoestrogen of chestnut-honey based propolis from Çarşamba district (Samsun, Turkey) by in silico molecular docking simulation.The present study evidenced the phenolic constituents from chestnut-honey based propolis possess phytoestrogen activity.