https://www.mdu.se/

Wearable sensors have garnered considerable interest from researchers in regard to their recent advancements. However, because of the pandemic and global turmoil, wearable devices have emerged as potential solution to combat the COVID-19 disease. Wearable devices enhanced with artificial intelligence (AI) can analyse extensive amounts of physiological data and provide users with estimates of possible risk factors and related illnesses. This review article focuses on the evolution and integration of AI in wearable sensors following pandemics. These technological breakthroughs have resulted in the development of contemporary wearable devices that are endowed with the ability to monitor various health parameters, anticipate prospective health issues, and provide early warnings before an event occurs. These advancements enhance our awareness for future health emergencies by enabling continuous monitoring and extensive data collection. To maximize the benefits of wearable technology for healthcare applications, it is very crucial to continuously develop sensor technology, enhance data integration, and establish robust regulatory frameworks.

Cardiovascular diseases are some of the underlying reasons contributing to the relentlessrise in mortality rates across the globe. In this regard, there is a genuine need to integrate advancedtechnologies into the medical realm to detect such diseases accurately. Moreover, numerous academicstudies have been published using AI-based methodologies because of their enhanced accuracy indetecting heart conditions. This research extensively delineates the different heart conditions, e.g.,coronary artery disease, arrhythmia, atherosclerosis, mitral valve prolapse/mitral regurgitation, andmyocardial infarction, and their underlying reasons and symptoms and subsequently introducesAI-based detection methodologies for precisely classifying such diseases. The review shows thatthe incorporation of artificial intelligence in detecting heart diseases exhibits enhanced accuraciesalong with a plethora of other benefits, like improved diagnostic accuracy, early detection and prevention,reduction in diagnostic errors, faster diagnosis, personalized treatment schedules, optimizedmonitoring and predictive analysis, improved efficiency, and scalability. Furthermore, the reviewalso indicates the conspicuous disparities between the results generated by previous algorithms andthe latest ones, paving the way for medical researchers to ascertain the accuracy of these resultsthrough comparative analysis with the practical conditions of patients. In conclusion, AI in heartdisease detection holds paramount significance and transformative potential to greatly enhancepatient outcomes, mitigate healthcare expenditure, and amplify the speed of diagnosis.

Cardiovascular diseases are some of the underlying reasons contributing to the relentless rise in mortality rates across the globe. In this regard, there is a genuine need to integrate advanced technologies into the medical realm to detect such diseases accurately. Moreover, numerous academic studies have been published using AI-based methodologies because of their enhanced accuracy in detecting heart conditions. This research extensively delineates the different heart conditions, e.g., coronary artery disease, arrhythmia, atherosclerosis, mitral valve prolapse/mitral regurgitation, and myocardial infarction, and their underlying reasons and symptoms and subsequently introduces AI-based detection methodologies for precisely classifying such diseases. The review shows that the incorporation of artificial intelligence in detecting heart diseases exhibits enhanced accuracies along with a plethora of other benefits, like improved diagnostic accuracy, early detection and prevention, reduction in diagnostic errors, faster diagnosis, personalized treatment schedules, optimized monitoring and predictive analysis, improved efficiency, and scalability. Furthermore, the review also indicates the conspicuous disparities between the results generated by previous algorithms and the latest ones, paving the way for medical researchers to ascertain the accuracy of these results through comparative analysis with the practical conditions of patients. In conclusion, AI in heart disease detection holds paramount significance and transformative potential to greatly enhance patient outcomes, mitigate healthcare expenditure, and amplify the speed of diagnosis.

Community-acquired pneumonia is one of the most lethal infectious diseases, especially for infants and the elderly. Given the variety of causative agents, the accurate early detection of pneumonia is an active research area. To the best of our knowledge, scoping reviews on diagnostic techniques for pneumonia are lacking. In this scoping review, three major electronic databases were searched and the resulting research was screened. We categorized these diagnostic techniques into four classes (i.e., lab-based methods, imaging-based techniques, acoustic-based techniques, and physiological-measurement-based techniques) and summarized their recent applications. Major research has been skewed towards imaging-based techniques, especially after COVID-19. Currently, chest X-rays and blood tests are the most common tools in the clinical setting to establish a diagnosis; however, there is a need to look for safe, non-invasive, and more rapid techniques for diagnosis. Recently, some non-invasive techniques based on wearable sensors achieved reasonable diagnostic accuracy that could open a new chapter for future applications. Consequently, further research and technology development are still needed for pneumonia diagnosis using non-invasive physiological parameters to attain a better point of care for pneumonia patients.

Machine learning has proven to be a powerful tool across many domains including healthcare. Heart sound classification using machine learning can revolutionize cardiac care by improving diagnostic accuracy, enabling early intervention, and facilitating personalized treatment strategies. However, obtaining labeled data for classification model training can be difficult, especially for rare or complex conditions. Furthermore, classifying heart sounds accurately can be challenging due to the variabilityin sound patterns and the presence of noise which requires preprocessing. In this paper, two machine learning models were trained and evaluated using DenseNet architecture on the CirCorDigiScope Phonocardiagram dataset and an ensemble of SupportVector Machine (SVM) and Decision Tree (DT) algorithms on acustom dataset curated by our partner, Tech4Life. The F1 score of the CirCor trained model was 75%. This is our effort to advance the application of machine learning in heart sound classification.

Speech emotion recognition (SER) is considered a pivotal area of research that holds significant importance in a variety of real-time applications, such as assessing human behavior and analyzing the emotional states of speakers in emergency situations. This paper assesses the capabilities of deep convolutional neural networks (CNNs) in this context. Both CNNs and Long Short-Term Memory (LSTM) based deep neural networks are evaluated for voice emotion identification. In our empirical evaluation, we utilize the Toronto Emotional Speech Set (TESS) database, which comprises speech samples from both young and old individuals, encompassing seven distinct emotions: anger, happiness, sadness, fear, surprise, disgust, and neutrality. To augment the dataset, variations in voice are introduced along with the addition of white noise. The empirical findings indicate that the CNN model outperforms existing studies on SER using the TESS corpus, yielding a noteworthy 21% improvement in average recognition accuracy. This work underscores SER’s significance and highlights the transformative potential of deep CNNs for enhancing its effectiveness in real-time applications, particularly in high-stakes emergency situations.

Photoplethysmography (PPG) signals offer a non-invasive and cost-effective mean sfor monitoring cardiovascular health. However, extracting clinically relevant information from these signals in real-time poses significant challenges. This paper presents a novel biosignal processingunit that utilizes the PPGFeat MATLAB toolbox to perform real-time signal processing and feature extraction from PPG signals, enabling continuous cardiovascular disease (CVD) monitoring andanalysis. We propose a system that interfaces with PPG sensors to acquire raw signals in real-time.The PPGFeat toolbox provides an interactive user interface, it identifies high-quality signals basedon their signal quality indices (SQIs) and performs segmentation The segmented PPG signals are then preprocessed by PPGFeat to remove noise and artifacts, smooth the waveforms, and correc tbaseline drift using a Chebyshev type II 4th order, 20 dB filter with a frequency range of 0.4–8 Hz.After preprocessing, a novel algorithm within PPGFeat is employed to accurately extract key fiducial points from the filtered PPG signals and their first and second derivatives. These includes ystolic peaks, diastolic peaks, onsets, and dicrotic notches, as well as inflection points, maxima, and minima on the derivative waveforms. Utilizing these extracted points, PPGFeat computes a comprehensive set of features, including pulse transit time, augmentation index, stiffness index,various magnitudes, and time intervals. These features characterize the PPG signal's morphology,timing intervals, and other relevant characteristics. These features are continuously streamed as output, providing a real-time stream of biomarkers and indicators for CVD analysis and monitoring.The resulting biomarkers and features can be fed into machine learning models or rule-based systems for real-time CVD identification, risk stratification, and monitoring applications. By utilizing PPGFeat's robust algorithms and proven accuracy, the proposed biosignal processing unit enables efficient real-time extraction of clinically relevant information from PPG signals, paving the way for improved cardiovascular health monitoring and personalized healthcare solutions.

Skin cancer is a rapidly growing and potentially deadly form of cancer, making early detection crucial for improved patient outcomes. This study introduces a machine learning-based educational system designed to aid early detection of skin cancer. The system leverages machine learning techniques to analyse skin lesion images from the ISIC-ISBI 2016 and 2017 datasets providing a non-invasive and cost-effective alternative to traditional biopsies. The primary objective of this dataset collection was to generate an automated prediction of lesion segmentation boundaries using dermoscopy images, where each image has a manual tracing of lesion boundaries done by an expert. To accommodate the diverse images acquired using many different devices, pre-processing and segmentation using OTSU thresholding isolate the region of interest, followed by extraction of detailed features such as texture, shape, and color. Principal component analysis (PCA) refines these features. An SVM ensemble classifier, trained on labeled images and evaluated on the ISIC-ISBI datasets, distinguishes cancerous from non-cancerous lesions. The system achieves an impressive95.73% accuracy, a 95.51% average similarity rate in segmentation, and a low mean squared error(MSE), demonstrating its effectiveness. This system operates in real-time as a user-friendly application executable on any desktop computer, tablet, or laptop. The application takes an image asinput, pre-processes it, and extracts relevant features. Using this feature matrix, the classifier determines whether the input image indicates a malignant or benign melanoma. The output provide sa clear label of 'cancerous melanoma' or 'benign melanoma' for each analyzed image. This system offers significant educational value for dermatology students and doctors. It can be used for hands-on learning and classroom training, enabling accurate diagnosis without the need for invasive biopsies. The system's potential portability makes it a valuable tool for resource-limited settings and large-scale educational initiatives focused on skin cancer detection.

The extraction of relevant features from the photoplethysmography signal for estimating certain physiological parameters is a challenging task. Various feature extraction methods have been proposed in the literature. In this study, we present a novel fiducial point extraction algorithm to detect c and d points from the acceleration photoplethysmogram (APG), namely “CnD”. The algorithm allows for the application of various pre-processing techniques, such as filtering, smoothing, and removing baseline drift; the possibility of calculating first, second, and third photoplethysmography derivatives; and the implementation of algorithms for detecting and highlighting APG fiducial points. An evaluation of the CnD indicated a high level of accuracy in the algorithm’s ability to identify fiducial points. Out of 438 APG fiducial c and d points, the algorithm accurately identified 434 points, resulting in an accuracy rate of 99%. This level of accuracy was consistent across all the test cases, with low error rates. These findings indicate that the algorithm has a high potential for use in practical applications as a reliable method for detecting fiducial points. Thereby, it provides a valuable new resource for researchers and healthcare professionals working in the analysis of photoplethysmography signals.

Photoplethysmography (PPG) has emerged as an increasingly attractive signal for non-invasive physiological measurements, owing to its simplicity, cost-effectiveness, and broad applicability spanning cardiovascular to respiratory systems. The burgeoning interest in PPG signal processing has facilitated its extensive incorporation in wearable devices, thus stimulating active research in this field. The present study undertakes a comprehensive evaluation to discern the optimal index finger (right or left) for PPG data acquisition and subsequent filtration, appraised through the lens of the signal-to-noise ratio (SNR) of the filtered signal. An analysis conducted on signals contaminated with white Gaussian noise unveiled that the Savitzky-Golay filter (a polynomial filter) with a window size of three outperformed other window lengths, rendering the highest SNR. Among the Infinite Impulse Response (IIR) filters compared; the Chebyshev I filter emerged as superior. Interestingly, the right index finger consistently demonstrated a higher mean SNR across filters: 0.49% for the Savitzky-Golay filters, 4.32% for the Butterworth (order 6), 7.71% for the Chebyshev I (order 10), and 4.02% for the Chebyshev II (order 4), relative to the left index finger for PPG signals perturbed by white Gaussian noise. These findings provide an insightful perspective for future research and development in wearable devices, suggesting potential superiority of the right index finger for PPG signal acquisition and filtration.