COVIR: A virtual rendering of a novel NN architecture O-Net for COVID-19 Ct-scan automatic lung lesions segmentation.

With the Coronavirus disease 2019 (COVID-19) spread, causing a world pandemic, and recently, the virus new variants continue to appear, making the situation more challenging and threatening, the visual assessment and quantification by expert radiologists have become costly and error-prone. Hence, there is a need to propose a model to predict the COVID-19 cases at the earliest possible to control the disease spread. In order to assist the medical professionals and reduce workload and the time the COVID-19 diagnosis cycle takes, this paper proposes a novel neural network architecture termed as O-Net to automatically segment chest Computerised Tomography Ct-scans infected by COVID-19 with optimised computing power and memory occupation. The O-Net consists of two convolutional autoencoders with an upsampling channel and a downsampling channel. Experimental tests show our proposal's effectiveness and potential, with a dice score of 0.86, pixel accuracy, precision, specificity of 0.99, 0.99, 0.98, respectively. Performance on the external dataset illustrates generalisation and scalability capabilities of the O-Net model to Ct-scan obtained from different scanners with different sizes. The second objective of this work is to introduce our virtual reality platform, COVIR, that visualises and manipulates 3D reconstructed lungs and segmented infected lesions caused by COVID-19. COVIR platform acts as a reading and visualisation support for medical practitioners to diagnose COVID-19 lung infection. The COVIR platform could be used for medical education professional practice and training. It was tested by Thirteen participants (medical staff, researchers, and collaborators), they conclude that the 3D VR visualisation of segmented Ct-Scan provides an aid diagnosis tool for better interpretation.

Classification of Fatigue Phases in Healthy and Diabetic Adults Using Wearable Sensor.

Fatigue is defined as "a loss of force-generating capacity" in a muscle that can intensify tremor. Tremor quantification can facilitate early detection of fatigue onset so that preventative or corrective controls can be taken to minimize work-related injuries and improve the performance of tasks that require high-levels of accuracy. We focused on developing a system that recognizes and classifies voluntary effort and detects phases of fatigue. The experiment was designed to extract and evaluate hand-tremor data during the performance of both rest and effort tasks. The data were collected from the wrist and finger of the participant's dominant hand. To investigate tremor, time, frequency domain features were extracted from the accelerometer signal for segments of 45 and 90 samples/window. Analysis using advanced signal processing and machine-learning techniques such as decision tree, k-nearest neighbor, support vector machine, and ensemble classifiers were applied to discover models to classify rest and effort tasks and the phases of fatigue. Evaluation of the classifier's performance was assessed based on various metrics using 5-fold cross-validation. The recognition of rest and effort tasks using an ensemble classifier based on the random subspace and window length of 45 samples was deemed to be the most accurate (96.1%). The highest accuracy (~98%) that distinguished between early and late fatigue phases was achieved using the same classifier and window length.

Covariance matrix based fall detection from multiple wearable sensors.

Falls are among the critical accidents experienced by elderly people and patients carrying some diseases. Subsequently, the detection and prevention of falls have become a hot research and industrial topic. This is due to the fact that falls are behind numerous irreversible injuries, or even death, and are weighting on the budgets of the health services. Automatic fall detection is one of the proposed solutions which aim at monitoring people who are likely to fall. Such solutions mitigate the fall impact by taking a quick action, e.g. in case of a fall occurrence, an alert is sent to the hospital. In this paper, we propose a new fall detection system relying on different signals acquired with multiple wearable sensors. Our system makes use of the covariance of the raw signals and the nearest neighbor classifier. Besides feature extraction, we also employ the covariance matrix as a straightforward mean for fusing signals from multiple sensors, to enhance the classification performance. Evaluation on two publicly available fall datasets, namely CogentLabs and DLR, demonstrates that the proposed approach is efficient when exploiting a single sensor as well as when fusing data from multiple sensors. Geodesic metrics are found to provide a higher fall detection accuracy than the Euclidean metric. The best obtained classification accuracies are 92.51% and 98.31% for CogentLabs and DLR datasets, respectively.