This study introduces a novel Facial Emotion Recognition (FER) algorithm designed for uncontrolled, or "in the wild," environments. Traditional image-based FER systems face challenges in these settings due to variations in pose, occlusion, lighting, and skin tones. Our approach overcomes these limitations by extracting and utilizing 3D facial landmarks instead of relying directly on image-based features. We employ the MediaPipe FaceMesh model to extract 478 normalized facial landmarks from the FER+ dataset, which includes images labeled with eight emotions. These landmarks define 2556 face tessellations, serving as embedding features for a transformer-based network. The algorithm's unique aspect lies in its ability to normalize images across different conditions and cameras, using a consistent set of features derived from 3D landmarks. This normalization enables the integration of diverse FER datasets for training, enhancing the algorithm’s applicability across various devices. Achieving 73.7% accuracy on the FER+ dataset, the algorithm demonstrates significant promise in emotion classification. Its adaptive attention mechanism focuses on well-represented landmarks, accounting for the face's angle and aspects, rather than solely on the emotion itself. This study marks a significant advancement in FER technology, offering a robust solution for emotion recognition in real-world and varied conditions.

In this paper, we propose human perception mechanism based CNN-RNN based approach to recognize dimensional emotion using Aff-Wild dataset. The proposed model recognizes dimensional expression through two steps. First, motivated by the human facial expression perception mechanism, the patches were extracted based on facial key points. Second, the facial expression was classified based on CNN-RNN model using extracted patches. The mean concordance correlation coefficient (CCC) of the proposed model is 0.322, indicating superior performance compared to the original CNN-RNN model, which achieved a mean CCC of 0.283.

Abstract As the pandemic situation continues, many people exercise at home. Mat Pilates is a popular workout and effective core strengthening. Although many researchers have conducted pose recognition studies for exercise posture correction, the study on Pilates exercise is only one case on static images. Therefore, for the purpose of exercise monitoring, we propose a real-time Pilates posture recognition system on a smartphone for exercise monitoring. We aimed to recognize 8 Pilates exercises—Bridge, Head roll-up, Hundred, Roll-up, Teaser, Plank, Thigh stretch, and Swan. First, the Blazepose model is used to extract body joint features. Then, we designed a deep neural network model that recognizes Pilates based on the extracted body features. It also measures the number of workouts, duration, and similarity to experts in video sequences. The precision, recall, and f1-score of the posture recognition model are 0.90, 0.87, and 0.88, respectively. The introduced application is expected to be used for exercise management at home.

Prediction of abdominal aortic aneurysm (AAA) growth is of essential importance for the early treatment and surgical intervention of AAA. Capturing key features of vascular growth, such as blood flow and intraluminal thrombus (ILT) accumulation play a crucial role in uncovering the intricated mechanism of vascular adaptation, which can ultimately enhance AAA growth prediction capabilities. However, local correlations between hemodynamic metrics, biological and morphological characteristics, and AAA growth rates present high inter-patient variability that results in that the temporal-spatial biochemical and mechanical processes are still not fully understood. Hence, this study aims to integrate the physics-based knowledge with deep learning with a patch-based convolutional neural network (CNN) approach by incorporating important multiphysical features relating to its pathogenesis for validating its impact on AAA growth prediction. For this task, we observe that the unstructured multiphysical features cannot be directly employed in the kernel-based CNN. To tackle this issue, we propose a parameterization of features to leverage the spatio-temporal relations between multiphysical features. The proposed architecture was tested on different combinations of four features including radius, intraluminal thrombus thickness, time-average wall shear stress, and growth rate from 54 patients with 5-fold cross-validation with two metrics, a root mean squared error (RMSE) and relative error (RE). We conduct extensive experiments on AAA patients, the results show the effect of leveraging multiphysical features and demonstrate the superiority of the presented architecture to previous state-of-the-art methods in AAA growth prediction.

To limit the spread of COVID-19, thermal screening cameras were installed everywhere. These cameras observe many thermal faces. These thermal face data are generally used to monitor strange temperatures for COVID-19 screening or to maintain social distancing. Big data of Thermal face generated everywhere should be used in the more practical functions. We proposed a method to measure non-contact breathing signals using thermal face data. In addition, breathing signals data estimated from thermal face data was converted to DICOM waveform Information Object Definitions (IODs) for interoperability management of medical data. The proposed method was tested on a golden reference (chest belt) with a mean accuracy of 93.52 %. a proposed method that can extract breathing signals using thermal screening cameras that are widely available around the world and manage data as healthcare interoperability information can show important potential in the public, telemedicine field in the future.

In the past few years, the use of big data has been important for public health. However, it was difficult to use medical imaging as public data. Several problems, such as data privacy and inconsistent measurement system, were not suitable for the interoperability in big data. In this paper, we proposed the reference model that satisfies the interoperability of medical imaging data and able to be shared in public data server. The de-identification method minimizes personal information, in order to protect the confidentiality of patients. The quantification method for medical imaging was to calibrate the measured values using a reference material. Finally, de-identified and quantified medical imaging data was shared with a public server.

Recently, portable electrocardiogram (ECG) hardware devices have been developed using limb-lead measurements. However, portable ECGs provide insufficient ECG information because of limitations in the number of leads and measurement positions. Therefore, in this study, V-lead ECG signals were synthesized from limb leads using an R-peak aligned generative adversarial network (GAN). The data used the Physikalisch-Technische Bundesanstalt (PTB) dataset provided by PhysioNet. First, R-peak alignment was performed to maintain the physiological information of the ECG. Second, time domain ECG was converted to bi-dimensional space by ordered time-sequence embedding. Finally, the GAN was learned through the pairs between the modified limb II (MLII) lead and each chest (V) lead. The result showed that the mean structural similarity index (SSIM) was 0.92, and the mean error rate of the percent mean square difference (PRD) of the chest leads was 7.21%.

In this paper, a method is proposed to measure human respiratory volume using a depth camera. The level-set segmentation method, combined with spatial and temporal information, was used to measure respiratory volume accurately. The shape of the human chest wall was used as spatial information. As temporal information, the segmentation result from the previous frame in the time-aligned depth image was used. The results of the proposed method were verified using a ventilator. The proposed method was also compared with other level-set methods. The result showed that the mean tidal volume error of the proposed method was 8.41% compared to the actual tidal volume. This was calculated to have less error than with two other methods: the level-set method with spatial information (14.34%) and the level-set method with temporal information (10.93%). The difference between these methods of tidal volume error was statistically significant [Formula: see text]. The intra-class correlation coefficient (ICC) of the respiratory volume waveform measured by a ventilator and by the proposed method was 0.893 on an average, while the ICC between the ventilator and the other methods were 0.837 and 0.879 on an average.

Three-dimensional (3-D) modeling and visualization of the cochlea using the World Wide Web (WWW) is an effective way of sharing anatomic information for cochlear implantation over the Internet, particularly for morphometry-based research and resident training in otolaryngology and neuroradiology. In this paper, 3-D modeling, visualization, and animation techniques are integrated in an interactive and platform-independent manner and implemented over the WWW. Cohen's template shape with mean cross-sectional areas of the human cochlea is extended into a 3-D geometrical model. Also, spiral computer tomography data of a patient's cochlea is digitally segmented and geometrically represented. The cochlear electrode array is synthesized according to its specification. Then, cochlear implantation is animated with both idealized and real cochlear models. Insertion length, angular position, and characteristic frequency of individual electrodes are estimated online during the virtual insertion. The optimization of the processing parameters is done to demonstrate the feasibility of this technology for clinical applications.