Traditional versus Neural Network Classification Methods for Facial Emotion Recognition

Facial emotion recognition (FER) is a topic that has gained interest over the years for its role in bridging the gap between Human and Machine interactions. This study explores the potential of real time FER modelling, to be integrated in a closed loop system, to help in treatment of children suffering from Autism Spectrum Disorder (ASD). The aim of this study is to show the differences between implementing Traditional machine learning and Deep learning approaches for FER modelling. Two classification approaches were taken, the first approach was based on classic machine learning techniques using Histogram of Oriented Gradients (HOG) for feature extraction, with a k-Nearest Neighbor and a Support Vector Machine model as classifiers. The second approach uses Transfer Learning based on the popular “Alex Net” Neural Network architecture. The performance of the approaches was based on the accuracy of randomly selected validation sets after training on random training sets of the Oulu-CASIA database. The data analyzed shows that traditional machine learning methods are as effective as deep neural net models and are a good compromise between accuracy, extracted features, computational speed and costs.

Comparison of biphasic material properties of equine articular cartilage estimated from stress relaxation and creep indentation tests

Mechanical parameters of hard and soft tissues are explicit markers for quantitative tissue characterization. In this study, we present a comparison of biphasic material properties of equine articular cartilage estimated from stress relaxation (ε = 6 %, t = 1000 s) and creep indentation tests (F = 0.1 N, t = 1000 s). A biphasic 3D-FE-based method is used to determine the biomechanical properties of equine articular cartilage. The FE-model computation was optimized by exploiting the axial symmetry and mesh resolution. Parameter identification was executed with the Levenberg- Marquardt-algorithm. Additionally, sensitivity analyses of the calculated biomechanical parameters were performed. Results show that the Young’s modulus E has the largest influence and the Poisson’s ratio of ν ≤ 0.1 is rather insensitive. The R² of the fit results varies between 0.882 and 0.974 (creep model) and between 0.695 and 0.930 (relaxation model). The averaged parameters E and k determined from the creep model yield higher values in comparison to the relaxation model. The differences can be traced back to the experimental settings and to the biphasic material model.

Validation of a Fluid-Structure Interaction Model for the Characterization of Transcatheter Mitral Valve Repair Devices

Mitral regurgitation (MR) is the second most frequent indication for heart valve surgery and catheter interventions. According to European and US-American guidelines, transcatheter mitral valve repair in general and transcatheter edge-to-edge repair (TEER) in particular may be considered as a treatment option for selected high-risk patients. However, the biomechanical impact of TEERdevices on the mitral valve (MV) has not yet been fully understood. To address this problem, a 3D-Fluid-Structure Interaction (FSI) framework utilizing non-linear Finite Element Analysis (FEA) for the MV apparatus and Smoothed Particle Hydrodynamics (SPH) for the pulsatile fluid flow was developed and validated against in vitro data. An artificial MV-model (MVM) with a prolapse in the A2-P2 region and a custom-made TEER device implanted in the A2-P2 region were used for the in vitro investigations. In accordance with ISO 5910, projected mitral orifice areas (PMOA), flow rates as well as atrial and ventricular pressures were measured under pulsatile flow conditions before and after TEER device implantation. For the FSI-model, the MVM geometry was reconstructed by means of microcomputed tomography in a quasi-stress-free configuration. Quasi-static tensile test data was utilized for the development of linear- and hyperelastic material models of the chordae tendineae and leaflets, respectively. The fluid flow was modelled assuming an incompressible, homogenous Newtonian behaviour. Time-varying in vitro transmitral pressure loading was applied as a boundary condition. In vitro investigations show that TEER device implantation in the A2-P2 region effectively reduces the regurgitation fraction (RF) from 55 % to 13 %. Moreover, the comparison of experimental and numerical data yields a deviation of 2.09 % for the RF and a deviation of 0.40 % and 6.47 % for the maximum and minimum PMOA, respectively. The developed FSI-framework is in good agreement with in vitro data and is therefore applicable for the characterization of the biomechanical impact of different TEER devices under pulsatile flow conditions.

Rapid Prototyping of Moulds for PDMS-based Microfluidic Chips

To shorten the production time for PDMS-moulds by additive manufacturing (AM) several 3D printers have been investigated in comparison to standard micro-milling by producing benchmark structures. These are evaluated regarding their shape accuracy, the transparency of the casted PDMS which is linked to the surface quality of the mould, and the production time until the moulds are ready to use. Even though the additively manufactured moulds showed significantly better surface quality and shorter production time, the necessary shape accuracy for non-square-structures or structures with < 250 μm edge length could not be achieved due to limiting factors like nozzle diameter, size of the digital micromirror device or spot size of the LCD-panel.

Partial 3D-reconstruction of the colon from monoscopic colonoscopy videos using shape-from-motion and deep learning

For the image-based documentation of a colonoscopy procedure, a 3D-reconstuction of the hollow colon structure from endoscopic video streams is desirable. To obtain this reconstruction, 3D information about the colon has to be extracted from monocular colonoscopy image sequences. This information can be provided by estimating depth through shape-from-motion approaches, using the image information from two successive image frames and the exact knowledge of their disparity. Nevertheless, during a standard colonoscopy the spatial offset between successive frames is continuously changing. Thus, in this work deep convolutional neural networks (DCNNs) are applied in order to obtain piecewise depth maps and point clouds of the colon. These pieces can then be fused for a partial 3D reconstruction.

Comparison of Characteristics of Poly(Nisopropylacrylamide) in Bulk Hydrogel and Ball-milled Microgel Forms

Poly(N-isopropylacrylamide) hydrogels are a popular temperature sensitive biomaterial. Bulk hydrogels can be quickly and easily processed into microgels using a ball mill. However, there is no information on whether the mechanical milling process affects critical PNIPAM characteristics. In this work, we compare swelling and thermo-responsive properties for a series of PNIPAM gels in bulk and microgel forms.

Finding the curved pathway of the large intestine for robot-aided colonoscopy

To assist the insertion of a robot-aided endoscope during colonoscopy, a measuring system is required so that the endoscope tip can align automatically and thus find the curved pathway of the large intestine. To achieve this, a selfexpanding nitinol wire basket is used to sense the contour of the intestine. As the wire basket touches the wall, it is deflected towards the center of the intestine. The relative position of the wire basket within the camera image is captured, which describes the desired direction to follow the organ. To identify the wire basket in the image, the original RGB image stream is converted into the HSV (hue, saturation, value) color space. Thus, a binary image can be created, in which only the neongreen color portion of the wire basket is visible as a cross. The Hough Transformation is used to search for straight lines in the binary image. Once two lines are found, the intersection point can be calculated and thus its position in the image. The evaluation of the execution time of the algorithm on a live stream was 45 ± 31 ms on average. The algorithm robustly recognizes the wire basket even if it was not visible to the human eye in the original RGB image due to deficient lighting.

Influence of pedicle screw thread width and recovery time after surgery on fixation strength

Introduction: Pedicle screw fixation systems are widely used for treatment of various spinal pathologies, including spinal stenosis, scoliosis, spinal deformities and fractures. Stress shielding is considered to be a major factor contributing to insufficient fixation strength, leading to screw loosening. In this study, the influence of pedicle screw thread width on the displacement of pedicle screw and stress transfer is analyzed using 2-Dimensional axisymmetric finite element (FE) model. Methods: FE model consisting of cancellous and cortical bone, along with pedicle screw is developed for this study. The pedicle screw thread width is varied between 0.1 mm and 0.6 mm in steps of 0.1 mm, while the other geometric parameters, including screw half-angle, pitch, diameter, and length are kept constant. Three different contact conditions between screw and bone, such as frictionless, frictional, and bonded are considered to simulate hours, days, and months after surgery, respectively. The material properties and boundary conditions are applied based on previous studies. An axial force of 80 N is applied on the screw head to simulate axial pull-out test. Results: Similar patterns of stress distribution are observed for all screw models, with high stress concentration above the first thread. The highest displacement in screw is observed shortly after surgery, while the highest displacement in cancellous and cortical bone is observed few days and months after the surgery, respectively. The average von Mises stress in screw decreases with increase in thread width for all contact conditions. In few hours/days after the surgery, stress transfer parameter increases with increase in thread width, up to a thread width of 0.5 mm and then decreases. The changes in stress transfer parameter are negligible few months after the surgery. Conclusion: This study highlights the influence of thread width on displacement and stress transferred to the bone, at different durations after the surgery. It is observed that a thread width of 0.5 mm exhibits the highest stress transfer, leading to reduced stress shielding and improved bone remodeling. It appears that this study might aid in developing better pedicle screws for the treatment of various spinal pathologies.

PTT-based Contact-less Blood Pressure Measurement using an RGB-Camera

Commonly used blood pressure measurement devices have noticeable limitations in accuracy, measuring time, comfort or safety. To overcome these limitations, we developed and tested a surrogate-based, non-invasive blood pressure measurement method using an RGB-camera. Our proposed method employs the relation between the pulse transit time (PTT) and blood pressure. Two remote photoplethysmography (rPPG) signals at different distances from the heart are extracted to calculate the temporal delay of the pulse wave. In order to establish the correlation between the PTT values and the blood pressure, a regression model is trained and evaluated. Tests were performed with five subjects, where each subject was recorded fifteen times for 30 seconds. Since the physiological parameters of the cardiac system are different for each person, an individual calibration is required to obtain the systolic and diastolic blood pressure from the PTT values. The calibration results are limited by the small number of samples and the accuracy of the reference system. However, our results show a strong correlation between the PTT values and the blood pressure and we obtained a mean error of 0.18 +/- 5.50 mmHg for the diastolic blood pressure and 0.01 +/- 7.71 mmHg for the systolic pressure, respectively.

Analysis of uterine electromyography signals using time-frequency based topological features for detection of preterm birth

Detection of preterm birth (gestational week < 37) is a global priority as it causes major health problems to neonates. Assessment of uterine contractions (burst) is required to detect and prevent the threat of preterm birth. Uterine electromyography (uEMG) is widely preferred to measure the uterine contractions noninvasively. These signals are nonstationary in nature. It can be handled by topological data analysis (TDA) effectively. Therefore, TDA can be used to explore the characteristics of uEMG burst signals. In this study, an attempt has been made to distinguish term (gestational week ≥ 37) and preterm conditions using timefrequency based topological features in uEMG burst signals. These signals are obtained from the publicly available online dataset. The annotated burst signals are segmented and subjected to a short time Fourier transform. The transformed real and imaginary Fourier coefficients are plotted in the complex plane and the envelope of the data points are computed using the alpha-shape technique. Four topological features such as, area, perimeter, circularity and ellipse variance are extracted. These features are statistically analyzed. The coefficient of variation (CoV) is calculated to measure the inter-subject variations. The results show that the proposed method is able to discriminate between term and preterm conditions. The extracted features namely, area and perimeter exhibit significant difference (p < 0.05) between these two conditions. The CoV of the perimeter is observed to be low, implying that this feature can handle inter-subject variations in burst signals. The extracted topological features are useful to analyze the characteristics of term and preterm pregnancies