Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference最新文献

Cardiopulmonary sounds contain a rich reservoir of vital and pathological information critical for clinical diagnosis. This paper presents a novel approach to cardiopulmonary data capturing with compressive sensing and reconstruction using a Convolutional Neural Network (CNN) based on the U-Net architecture. Applying traditional compressive sensing techniques to cardiopulmonary sounds presents several challenges. Cardiopulmonary sounds are inherently complex, with a substantial variation between captures. The traditional algorithms for compressive sensing rely on signal sparsity, whereas finding a sparse representation domain for cardiopulmonary sounds is a difficult task. Instead of finding a sparse domain manually, we propose training a convolutional encoder-decoder neural network for a pseudo-randomly undersampled set of signals without explicitly enforcing the sparsity concept. In this research, a CNN was trained for pseudo-randomly decimated input signals, evaluating a compression ratio of up to 30. The network is trained for respiratory sounds using the SPRSound dataset and for Phonocardiogram (PCG) signals using the CirCor Digiscope PCG dataset. Both these datasets have been evaluated for signal integrity after reconstruction and delivered promising results. The algorithm achieves reconstruction quality similar to that of previous research with a compression ratio three times higher than that of previous research applied to respiratory sounds. Since the principles of compressive sensing are applied in the sampling stage, the data compression requires no computation in the compression stage, and can therefore easily be implemented in low-cost edge devices.Clinical relevance- This work enables efficient compression of cardiopulmonary sounds, maintaining high signal integrity even at three times higher compression ratios than previous methods applied to respiratory sounds. It supports low-power, portable devices for real-time monitoring, improving accessibility for telemedicine and point-of-care diagnostics in respiratory and cardiovascular conditions.

Atrial fibrillation (AF) is a common cardiac arrhythmia causing severe complications if left untreated. Due to its sporadic nature, early detection often requires longitudinal ambulatory electrocardiogram (ECG) screening. Recently, deep learning (DL) has gained prominence in analysing long-term ECG and automating AF detection. However, like any medical classification problem, obtaining diverse labelled ECG data for DL model training is expensive and time-consuming. This paper proposes a semi-supervised learning (SSL) based AF detection model employing a variational auto-encoder (VAE). It leverages varying amounts of labelled and unlabelled ECG data to optimise the AF detection performance on ambulatory ECG. As ambulatory contexts under free-living conditions influence ECG recordings, we incorporate context via accelerometry data and experiment with its influence on model performance. The proposed SSL model was trained on ECG data from 72,003 unique patients and can classify between sinus rhythms, AF, and other arrhythmias. Experimental results on unseen test dataset and the publicly available CACHET-CADB dataset clearly demonstrate the model's generalisability, achieving an accuracy of over 91% with just 20% of the training set being labelled. With extensive experiments, our study exhibits the ability of SSL to improve AF detection from ambulatory ECG using small amounts of labelled data.

Parkinson's disease (PD) is a neurodegenerative disorder marked by motor symptoms such as bradykinesia, tremor, rigidity, and postural instability, which impair gait and balance. As PD progresses, gait disturbances-including reduced speed, shorter strides, and freezing of gait (FoG)-increase the risk of falls and limit functional independence. While wearable sensors are commonly used to monitor gait in PD, there has been limited research on technological devices designed to assist mobility in this population. This study explored the feasibility of mCrutch, a sensorized crutch system, in supporting gait in individuals with PD. Participants wore 7 inertial measurement units. They completed a 2-minute walking task, clinical scales, and a survey to get feedback using mCrutch. This observational study is designed to: (i) explore the feasibility and acceptance of using the mCrutch system in people with PD and (ii) investigate whether clinical and gait parameters are related to mCrutch use during walking. Preliminary results indicated high user satisfaction, supporting the feasibility of mCrutch in clinical settings. Preliminary observations among the five participants suggest a potential correlation between mCrutch usage and gait speed and cadence. Additionally, mCrutch metrics may be associated with balance and clinical scales, particularly MDS-UPDRS scores, suggesting that higher disease severity corresponds to greater reliance on the device. Future work will focus on expanding the sample size to validate these preliminary findings.Clinical Relevance- This study preliminarily shows the potential of sensorized assistive devices like mCrutch to monitor and assist mobility in individuals with idiopathic PD.

Hand mobility impairments caused by conditions such as stroke, spinal cord injuries, and neuromuscular disorders significantly affect an individual's ability to perform daily activities. This paper presents a Adaptive Hand Mobility System (AHMS) that utilizes flex sensors and motorized actuation to detect and amplify slight finger movements, facilitating precise hand motions for both assistance and rehabilitation. The system is wearable, lightweight, and wireless, providing real-time feedback through a sensor-driven adaptive control mechanism. Unlike conventional rehabilitation devices, which are often bulky and limited in functionality, this device offers multiple control modes, including manual assistance, cyclic movement training, and task-specific rehabilitation routines. A companion mobile application integrates predefined physiotherapy exercises and interactive therapeutic games, allowing users to engage in customized rehabilitation programs. Experimental evaluations demonstrate the system's effectiveness in enhancing grip strength, dexterity, and fine motor control, making it a promising solution for personalized rehabilitation and assistive mobility technology.

Prosthetic vision has emerged as a promising solution for restoring sight in visually impaired individuals. However, suboptimal perceptions prevent users from performing daily tasks effectively. Recent studies have shown that both physical constraints and anatomical characteristics contribute to phosphene distortions, highlighting the need for a personalized approach to enhance user experience. In this context, integrating deep learning-based strategies with prosthetic models and patient-specific information has demonstrated strong potential in generating more useful perceptions. Our approach improves upon previous methods by introducing a novel neural network architecture that incorporates a vision transformer to analyze both visual input and patient-specific parameters, aiming to reduce distortions through optimized stimulation parameters. Additionally, we develop geometric transformations to correct rotations and translations within the implant's field of view. The proposed model outperforms baseline methods on the MNIST dataset and sets a new baseline for more complex images, generating suitable perceptions for classification tasks in ImageNet, CIFAR-10 and COCO datasets.

Skin lesion classification using deep learning techniques is challenged by insufficient samples and class imbalances in datasets. This study introduces a novel framework, the class expert Deep Convolutional Generative Adversarial Network (DCGAN), designed to handle class imbalance and enhance classification accuracy for under represented classes. The proposed framework also leverages weight transfer from the GAN discriminator trained on each class to expert layers, which are then modified to classify skin lesion images more accurately using the discriminator's weights. This transfer learning strategy enhances the performance of the Convolutional Neural Network (CNN) model in DCGAN by utilizing the discriminative features learned during GAN training. Experimental evaluations demonstrate that the proposed class expert DCGAN framework achieves notable improvements in accuracy and precision, particularly for classes with fewer samples. Specifically, it achieves a 2-3% increase in classification accuracy compared to traditional methods. These results underscore the effectiveness of leveraging GANs for data augmentation and discriminative feature extraction in medical image classification. Thus, the class expert DCGAN framework offers a promising solution to improve the performance of skin lesion classification models, facilitating highly reliable diagnostic decisions and enhancing the interpretation of dermatological images across diverse clinical scenarios.

Global warming is one of the most relevant effects of climate change, and poses a serious hazard for human health, in particular in relation to the cardiovascular system, leading to an increased short-term risk of Out-of-Hospital Cardiac Arrest (OHCA). This study examines this risk increase from a geospatial viewpoint, going beyond pathophysiology, and emphasizing the need for a public health-focused, multidisciplinary approach known as environmental epidemiology. While some solutions have already been proposed (in particular, risk indexing as defined by the Intergovernmental Panel on Climate Change, IPCC, and the Distributed Lag Non-linear Model, DLNM), they require complex and manifold data (thus limiting replicability), are computationally intensive, and cannot be easily interpreted. To address these gaps, this research introduces a Geospatial Heat-related Risk Index (GHRI) for territorial risk stratification, aiding in efficient Emergency Medical Services (EMS) resource planning. Focusing on Lombardy, Italy, a densely populated region with diverse climates, the study analyzes temperature data from the Regional Agency for Environmental Protection and OHCA records from the Regional Agency for Emergency/Urgency (AREU) between 2017 and 2021. Data were mapped onto 96 Base Statistical Areas (BSAs), each with approximately 100'000 residents. Using Geographic Information Systems (GIS) and Python, the study finds that heat exposure generally increases OHCA risk, though some areas showed protective or insignificant effects.. The findings highlight the importance of GIS-based environmental epidemiology in climate adaptation policies and emergency healthcare planning, providing actionable insights for public health strategies.Clinical Relevance- The proposed framework allows to identify territories that exhibit higher risk in terms of increased out-of-hospital cardiac arrest incidence during heat days, thus providing valuable information to support planning and management of Emergency Medical Services (EMS). More efficient resources allocation reduces intervention time and increases patients' survival probability, which is particularly critical for out-of-hospital cardiac arrest.

We test the hypothesis that amplitude permutation conditional entropy (APCE) is more powerful than permutation conditional entropy (PCE) when complexity of heart period (HP) dynamics is decreased by vagal blockade or withdrawal. We acquired HP variability in 9 healthy male physicians (age: 25-46 yrs) at baseline (B) and during administration of a high dose of atropine (AT) and in 15 healthy nonsmoking volunteers (age: 24-54 yrs, 9 males and 6 females) at rest in horizontal position (T0) and during 90° head-up tilt (T90). In addition to coarse-graining-free methods, like PCE and APCE, we computed coarse-graining-based k-nearest-neighbor conditional entropy (KNNCE) for comparison. Markers were computed over 256 consecutive HP values, thus targeting the complexity of short-term cardiac control. PCE was unable to detect the decrease of HP variability complexity during AT compared to B, while APCE and KNNCE could. All the conditional entropy markers found a decrease in HP variability complexity during T90 compared to T0. Only APCE was correlated with KNNCE in both protocols. We conclude that APCE is more reliable than PCE in assessing cardiac control complexity, likely due to the better ability of APCE in the presence of the low signal-to-noise ratio of HP dynamics observed during AT.

The rising global prevalence of heart disease necessitates early detection for improved diagnosis and treatments. Automated echocardiography robotic systems are revolutionizing cardiology by enhancing diagnostic accuracy and efficiency. These systems integrate real-time image acquisition and processing to navigate patient anatomy and adapt imaging techniques dynamically without human intervention. Accurate cardiac view classification is vital for capturing diagnostically relevant images, forming the basis for subsequent automated disease detection and diagnosis. Although deep learning has emerged as a powerful tool for medical image analysis, its application in echocardiography remains limited due to the complexity of multi-view echocardiography imaging. The proposed system leverages deep learning models, specifically convolutional neural networks, trained on a diverse dataset of echocardiographic images to distinguish standard cardiac views, including the parasternal long-axis, parasternal short-axis, and apical four-chamber views. This capability enables the robotic system to autonomously navigate patient anatomy and optimize image acquisition in real time, minimizing operator dependency and ensuring imaging consistency. The long-term objective of this study is to develop a fully autonomous robotic system capable of early and accurate cardiovascular disease diagnosis, ultimately reducing diagnostic delays and improving patient outcomes.

This study aims to investigate whether quantitative radiomic features extracted from Positron Emission Tomography/Computed Tomography (PET/CT) could differentiate triple-negative breast cancer (TNBC) from non-triple-negative breast cancer (non-TNBC). We propose a pipeline that combines deep learning for cancer lesion segmentation with machine learning techniques to classify TNBC. Our approach leveraged the radiomic features extracted from ¹⁸F-fluorodeoxyglucose PET/CT. This retrospective study included the PET/CT images of 217 patients with breast cancer (57 TNBC and 160 non-TNBC) admitted to Georges-François Leclerc Hospital. The tumor regions of interest were automatically segmented on PET images using a deep learning model and mapped to CT scans. Radiomic features were extracted from 3D tumor volumes and machine learning classifiers were built using stratified 5-fold cross-validation. Recursive feature elimination was employed to rank and select the most relevant radiomic features, thereby enhancing classification performance. The model was evaluated using the F1-score, area under the receiver operating characteristic (ROC) curve (AUC), accuracy, sensitivity, and specificity. The proposed method achieved promising performance, with an F1-score of 0.90 ± 0.02, an accuracy of 0.86 ± 0.07, a sensitivity of 0.91 ± 0.06, and an AUC of 0.88 ± 0.04, using the top-ranked features. The metrics were evaluated as the average over a five-fold cross-validation. Radiomic features extracted from PET and CT scans provide valuable prognostic insights for the identification of TNBC. This study demonstrated that machine learning algorithms based on radiomic features and automated PET/CT segmentation can accurately distinguish TNBC from non-TNBC.Clinical relevance- This study demonstrates the potential of image-based radiomic analysis combined with machine learning to differentiate triple-negative breast cancer (TNBC) from non-TNBC. By using deep learning for automatic tumor segmentation and feature extraction, this approach offers a non-invasive, quantitative tool that can improve TNBC diagnosis and the efficiency of treatment strategies. These advancements may help clinicians provide more reliable insights, while reducing the likelihood of misclassification.