Machine Learning for Human Motion Intention Detection.

The gait pattern of exoskeleton control conflicting with the human operator's (the pilot) intention may cause awkward maneuvering or even injury. Therefore, it has been the focus of many studies to help decide the proper gait operation. However, the timing for the recognization plays a crucial role in the operation. The delayed detection of the pilot's intent can be equally undesirable to the exoskeleton operation. Instead of recognizing the motion, this study examines the possibility of identifying the transition between gaits to achieve in-time detection. This study used the data from IMU sensors for future mobile applications. Furthermore, we tested using two machine learning networks: a linearfFeedforward neural network and a long short-term memory network. The gait data are from five subjects for training and testing. The study results show that: 1. The network can successfully separate the transition period from the motion periods. 2. The detection of gait change from walking to sitting can be as fast as 0.17 s, which is adequate for future control applications. However, detecting the transition from standing to walking can take as long as 1.2 s. 3. This study also find that the network trained for one person can also detect movement changes for different persons without deteriorating the performance.

A novel sound-blocking structure based on the muffler principle for rib-sparing transcostal high-intensity focused ultrasound treatment.

The main challenge in transcostal high-intensity focused ultrasound therapy is minimising heat deposition in the ribs while ensuring that a sufficient dose is delivered to the target region. Current approaches rely on expensive multichannel phased-array systems to turn the individual transducer on and off according to either geometrical arrangements or complicated wave calculations. To protect the ribs from heating, the ultrasound energy must not only not reach the ribs, but must also not accumulate in front of the ribs. The research in this paper proposes a different approach, of attaching a sound-blocking structure in front of the rib cage with similar effects to those of an engine exhaust muffler. The sound-blocking structure is based on the muffler principle to prevent ultrasound energy from reaching the ribs and reduce the amount of energy reflected back to the applicator. Finite element simulations with a 0.5-MHz transducer of the overall sound fields and temperature distribution showed that the ultrasound pressure and energy level would decrease behind the novel sound-blocking structures, thereby resulting in a lower temperature at the ribs than at the tumour. Without the protecting structure, the rib temperature reached 104.19 °C whereas with the structure it reached only 37.86 °C. An experimental set-up using porcine ribs with a phantom was also developed to validate the concept, which showed that the rib temperature reached 73 °C without protection within 1 min of ablation time whereas it reached 36.5 °C with the device. The tumour region in the tests reached 51 °C and 49 °C with and without protection, respectively.

A corneal elastic dynamic model derived from Scheimpflug imaging technology.

PURPOSE: To simultaneously extract the corneal Young's modulus and the damping ratio from Scheimpflug imaging data. METHODS: A spherical diaphragm model can better represent the geometry and physics of an eyeball than the popular mass-spring-damper model. This research derived the dynamic model of a water-filled spherical diaphragm based on the hydrodynamics and wave propagation theories. By applying modal analysis on the model, one can decouple the cornea vibration into individual modes and reconstruct the air puff vibration from the decoupled responses. By matching this response with the Scheimpflug imaging data from the Corvis(®) ST, it was then possible to extract multiple physiological properties as desired. RESULTS: The dynamic modal analysis was employed to extract the corneal physiological properties of 25 Taiwanese normal subjects. Specifically, the corneal Young's moduli and damping ratios were estimated. In fact the model is dependent on the physiological parameters such as cornea thickness, densities, and intraocular pressure. It is thus also possible to extract these parameters through multi-goal minimisation processes. CONCLUSIONS: The spherical diaphragm model was able to better describe the dynamic response of the eyeball. The model analysis also provides additional corneal physiological properties that were not available through other means.

Temperature effects on the magnetic properties of silicon-steel sheets using standardized toroidal frame.

This study designed a detachable and standardized toroidal test frame to measure the electromagnetic characteristic of toroidal laminated silicon steel specimens. The purpose of the design was to provide the measurements with standardized and controlled environment. The device also can withstand high temperatures (25-300°C) for short time period to allow high temperature tests. The accompanying driving circuit facilitates testing for high frequency (50-5,000 Hz) and high magnetic flux (0.2-1.8 T) conditions and produces both sinusoidal and nonsinusoidal test waveforms. The thickness of the stacked laminated silicon-steel sheets must be 30~31 mm, with an internal diameter of 72 mm and an outer diameter of 90 mm. With the standardized setup, it is possible to carry out tests for toroidal specimen in high temperature and high flux operation. The test results show that there is a tendency of increased iron loss under high temperature operation. The test results with various driving waveforms also provide references to the required consideration in engineering designs.

A robust variable sampling time BLDC motor control design based upon µ-synthesis.

The variable sampling rate system is encountered in many applications. When the speed information is derived from the position marks along the trajectory, one would have a speed dependent sampling rate system. The conventional fixed or multisampling rate system theory may not work in these cases because the system dynamics include the uncertainties which resulted from the variable sampling rate. This paper derived a convenient expression for the speed dependent sampling rate system. The varying sampling rate effect is then translated into multiplicative uncertainties to the system. The design then uses the popular µ-synthesis process to achieve a robust performance controller design. The implementation on a BLDC motor demonstrates the effectiveness of the design approach.

Human Posture Transition-Time Detection Based upon Inertial Measurement Unit and Long Short-Term Memory Neural Networks.

As human-robot interaction becomes more prevalent in industrial and clinical settings, detecting changes in human posture has become increasingly crucial. While recognizing human actions has been extensively studied, the transition between different postures or movements has been largely overlooked. This study explores using two deep-learning methods, the linear Feedforward Neural Network (FNN) and Long Short-Term Memory (LSTM), to detect changes in human posture among three different movements: standing, walking, and sitting. To explore the possibility of rapid posture-change detection upon human intention, the authors introduced transition stages as distinct features for the identification. During the experiment, the subject wore an inertial measurement unit (IMU) on their right leg to measure joint parameters. The measurement data were used to train the two machine learning networks, and their performances were tested. This study also examined the effect of the sampling rates on the LSTM network. The results indicate that both methods achieved high detection accuracies. Still, the LSTM model outperformed the FNN in terms of speed and accuracy, achieving 91% and 95% accuracy for data sampled at 25 Hz and 100 Hz, respectively. Additionally, the network trained for one test subject was able to detect posture changes in other subjects, demonstrating the feasibility of personalized or generalized deep learning models for detecting human intentions. The accuracies for posture transition time and identification at a sampling rate of 100 Hz were 0.17 s and 94.44%, respectively. In summary, this study achieved some good outcomes and laid a crucial foundation for the engineering application of digital twins, exoskeletons, and human intention control.

Effectiveness of external respiratory surrogates for in vivo liver motion estimation.

PURPOSE: Due to low frame rate of MRI and high radiation damage from fluoroscopy and CT, liver motion estimation using external respiratory surrogate signals seems to be a better approach to track liver motion in real-time for liver tumor treatments in radiotherapy and thermotherapy. This work proposes a liver motion estimation method based on external respiratory surrogate signals. Animal experiments are also conducted to investigate related issues, such as the sensor arrangement, multisensor fusion, and the effective time period. METHODS: Liver motion and abdominal motion are both induced by respiration and are proved to be highly correlated. Contrary to the difficult direct measurement of the liver motion, the abdominal motion can be easily accessed. Based on this idea, our study is split into the model-fitting stage and the motion estimation stage. In the first stage, the correlation between the surrogates and the liver motion is studied and established via linear regression method. In the second stage, the liver motion is estimated by the surrogate signals with the correlation model. Animal experiments on cases of single surrogate signal, multisurrogate signals, and long-term surrogate signals are conducted and discussed to verify the practical use of this approach. RESULTS: The results show that the best single sensor location is at the middle of the upper abdomen, while multisurrogate models are generally better than the single ones. The estimation error is reduced from 0.6 mm for the single surrogate models to 0.4 mm for the multisurrogate models. The long-term validity of the estimation models is quite satisfactory within the period of 10 min with the estimation error less than 1.4 mm. CONCLUSIONS: External respiratory surrogate signals from the abdomen motion produces good performance for liver motion estimation in real-time. Multisurrogate signals enhance estimation accuracy, and the estimation model can maintain its accuracy for at least 10 min. This approach can be used in practical applications such as the liver tumor treatment in radiotherapy and thermotherapy.

Correlation between corneal dynamic responses and keratoconus topographic parameters.

OBJECTIVE: To investigate the correlation between corneal biomechanical properties and topographic parameters using machine learning networks for automatic severity diagnosis and reference benchmark construction. METHODS: This was a retrospective study involving 31 eyes from 31 patients with keratonus. Two clustering approaches were used (i.e., shape-based and feature-based). The shape-based method used a keratoconus benchmark validated for indicating the severity of keratoconus. The feature-based method extracted imperative features for clustering analysis. RESULTS: There were strong correlations between the symmetric modes and the keratoconus severity and between the asymmetric modes and the location of the weak centroid. The Pearson product-moment correlation coefficient (PPMC) between the symmetric mode and normality was 0.92 and between the asymmetric mode and the weak centroid value was 0.75. CONCLUSION: This study confirmed that there is a relationship between the keratoconus signs obtained from topography and the corneal dynamic behaviour captured by the Corvis ST device. Further studies are required to gather more patient data to establish a more extensive database for validation.

Functional assistance for stress distribution in cell culture membrane under periodically stretching.

Dynamic cell cultures simulate the in vivo cell environment for a regular loading system with curtain strains. However, it is difficult to obtain strains that are suitable for cells without conducting multiple trials. This study develops a device that increases the strain gradient by changing the tensile section, in order to determine the effect of various cyclic strains on cultured human keratinocytes (HK) cells. This device is used to determine the effect of 3% and 5% cyclic strain and shear strain on cell proliferation and arrangement at 1 Hz. The results show that compared with static and 3% strain, a 5% cyclic strain better inhibits the proliferation of HK cells. Compared to the initial cell attachment when there is no specific directionality, the cells are aligned in the vertical stretching direction after cyclic stretching. This equipment increases the efficiency of the experiment and more intuitively maps the cell behavior and shape to the strain field and the response to the shear strain.

Changes in Intraocular Pressure after Transepithelial Photorefractive Keratectomy and Femtosecond Laser In Situ Keratomileusis.

PURPOSE: To investigate the changes in intraocular pressure (IOP) and biomechanically corrected IOP (bIOP) in patients undergoing transepithelial photorefractive keratectomy (TPRK) and femtosecond laser in situ keratomileusis (FS-LASIK) and to determine the effects of preoperative biomechanical factors on IOP and bIOP changes after FS-LASIK and TPRK. DESIGN: A retrospective comparative study. METHODS: We retrospectively investigated the IOP and corneal biomechanical changes in 93 eyes undergoing FS-LASIK and 104 eyes undergoing TPRK in a clinical setting. Preoperative and postoperative data on ophthalmic and Corvis ST examinations, in vivo Young's modulus, and noncontact tonometry were analyzed. Marginal linear regression models with generalized estimating equations were used for intragroup and intergroup comparisons of IOP and bIOP changes. RESULTS: In the univariate model, IOP reduction after FS-LASIK was 2.49 mmHg higher than that after TPRK. In addition, bIOP reduction after FS-LASIK was 1.85 mmHg higher than that after TPRK. In the multiple regression model, we revealed that IOP reduction after FS-LASIK was 1.75 mmHg higher than that after TPRK. Additionally, bIOP reduction after FS-LASIK was 1.64 mmHg higher than that after TPRK. Postoperative changes in bIOP were less than those in IOP. In addition, Young's modulus and CBI had no significant effect on postoperative IOP and bIOP changes. We establish a biomechanically predictive model using the available data to predict postoperative IOP and bIOP changes after TPRK and FS-LASIK. CONCLUSIONS: Reductions in IOP and bIOP after FS-LASIK were 1.75 mmHg and 1.64 mmHg, respectively, more than those after TPRK, after adjustment for confounders. We revealed that the type of refractive surgery and peak distance (PD) were significant predictors of postoperative IOP and bIOP changes. By contrast, depth of ablation showed a significant effect on only IOP changes.

Tracking control of shape-memory-alloy actuators based on self-sensing feedback and inverse hysteresis compensation.

Shape memory alloys (SMAs) offer a high power-to-weight ratio, large recovery strain, and low driving voltages, and have thus attracted considerable research attention. The difficulty of controlling SMA actuators arises from their highly nonlinear hysteresis and temperature dependence. This paper describes a combination of self-sensing and model-based control, where the model includes both the major and minor hysteresis loops as well as the thermodynamics effects. The self-sensing algorithm uses only the power width modulation (PWM) signal and requires no heavy equipment. The method can achieve high-accuracy servo control and is especially suitable for miniaturized applications.

Keratoconus Screening Based on Deep Learning Approach of Corneal Topography.

Purpose: To develop and compare deep learning (DL) algorithms to detect keratoconus on the basis of corneal topography and validate with visualization methods. Methods: We retrospectively collected corneal topographies of the study group with clinically manifested keratoconus and the control group with regular astigmatism. All images were divided into training and test datasets. We adopted three convolutional neural network (CNN) models for learning. The test dataset was applied to analyze the performance of the three models. In addition, for better discrimination and understanding, we displayed the pixel-wise discriminative features and class-discriminative heat map of diopter images for visualization. Results: Overall, 170 keratoconus, 28 subclinical keratoconus and 156 normal topographic pictures were collected. The convergence of accuracy and loss for the training and test datasets after training revealed no overfitting in all three CNN models. The sensitivity and specificity of all CNN models were over 0.90, and the area under the receiver operating characteristic curve reached 0.995 in the ResNet152 model. The pixel-wise discriminative features and the heat map of the prediction layer in the VGG16 model both revealed it focused on the largest gradient difference of topographic maps, which was corresponding to the diagnostic clues of ophthalmologists. The subclinical keratoconus was positively predicted with our model and also correlated with topographic indexes. Conclusions: The DL models had fair accuracy for keratoconus screening based on corneal topographic images. The visualization mentioned in the current study revealed that the model focused on the appropriate region for diagnosis and rendered clinical explainability of deep learning more acceptable. Translational Relevance: These high accuracy CNN models can aid ophthalmologists in keratoconus screening with color-coded corneal topography maps.

Facile fabrication of superporous and biocompatible hydrogel scaffolds for artificial corneal periphery.

Biocompatible and highly porous network hydrogel scaffolds were fabricated for the development of artificial cornea (AC) periphery/skirt that could be used to enhance the long-term retention of the implants. In this study, a series of hydrogel scaffolds for this application was fabricated from the photo-polymerization of a mixture of poly(ethylene glycol) (PEG)- and poloxamer (P407)-based macromer solutions in dichloromethane in which solvent-induced phase separation (SIPS) arose to form scaffolds with macroporous structure and high water content. The overall porosity ranging from 20% to 75% and open/closed pore structure of the hydrogel scaffolds could be finely tuned by varying the ratio of P407/PEG in the macromer solution and solvent type. The total porosity and open-cell structure of the macropores in the synthesized hydrogel scaffolds affected the swelling behavior, dynamic properties such as the storage moduli of the hydrogels as well as their degradation rates. Based on the subcutaneous implantation in rats, superporous hydrogel scaffolds induced the formation of thinner fibrous capsules around the implants and showed less inflammatory reaction, suggesting that the hydrogel scaffolds made from SIPS exhibited good cytocompatibility. The combined results of swelling ratio, porosity, physical strength and subcutaneous implant tests indicated that the superporous hydrogels with porosity >50% showed potentials to be used for cornea periphery application.

Minimizing abdominal wall damage during high-intensity focused ultrasound ablation by inducing artificial ascites.

High-intensity focused ultrasound (HIFU) is becoming an important tool for tumor treatment [especially hepatocellular carcinoma (HCC)] in Asian countries. A HIFU system provides unique advantages of low invasiveness and absence of nonradiation. However, if the target HCC is close to the proximal surface of the liver, HIFU may overheat diaphragm, abdominal wall or skin. To avoid this complication, a method using artificial ascites in the abdominal cavity to separate the liver from the peritoneum, and to serve as a heat sink to cool overlying structures and thereby avoid inducing permanent damage was proposed. Target tissue that was 10 mm below the liver surface was ablated in 12 New Zealand white rabbits: 6 in the experimental group and 6 in the control group. Artificial ascites was established in the experimental group by injecting normal saline into the abdominal cavity until the pressure reached 150 mm H2O. Artificial ascites not only reduced the probability and extent of thermal damage to intervening structures, but also had no adverse affect on the efficacy of HIFU ablation.

Biomechanical Simulation of Stress Concentration and Intraocular Pressure in Corneas Subjected to Myopic Refractive Surgical Procedures.

Recent advances in the analysis of corneal biomechanical properties remain difficult to predict the structural stability before and after refractive surgery. In this regard, we applied the finite element method (FEM) to determine the roles of the Bowman's membrane, stroma, and Descemet's membrane in the hoop stresses of cornea, under tension (physiological) and bending (nonphysiological), for patients who undergo radial keratotomy (RK), photorefractive keratectomy (PRK), laser-assisted in situ keratomileusis (LASIK), or small incision lenticule extraction (SMILE). The stress concentration maps, potential creak zones, and potential errors in intraocular pressure (IOP) measurements were further determined. Our results confirmed that the Bowman's membrane and Descemet's membrane accounted for 20% of the bending rigidity of the cornea, and became the force pair dominating the bending behaviour of the cornea, the high stress in the distribution map, and a stretch to avoid structural failure. In addition, PRK broke the central linking of hoop stresses and concentrated stress on the edge of the Bowman's membrane around ablation, which posed considerable risk of potential creaks. Compared with SMILE, LASIK had a higher risk of developing creaks around the ablation in the stroma layer. Our FEM models also predicted the postoperative IOPs precisely in a conditional manner.

Estimation of the Corneal Young's Modulus In Vivo Based on a Fluid-Filled Spherical-Shell Model with Scheimpflug Imaging.

Current intraocular pressure (IOP) measurement using air puff could be erroneous without applying proper corrections. Although noncontact tonometry is not considered to be accurate, it is still popularly used by eye clinics. It is thus necessary to extract the correct information from their results. This study proposes a practical approach to correctly measure IOP in vivo. By embedding a new model-based correction to the Corvis® ST, we can extract the corneal Young's modulus from the patient data. This Young's modulus can be used to correct the IOP readings. The tests were applied to 536 right eyes of 536 healthy subjects (228 male and 308 female) between March of 2012 and April of 2016. The tests were applied to patients at the Department of Ophthalmology, National Taiwan University Hospital and the Hung-Chuo Eye Clinics. The statistical analysis showed that the value for the Young's modulus was independent of all the other parameters collected from the Corvis ST, including the corneal thickness and the intraocular pressure. Therefore, it is important to independently measure the Young's modulus instead of depending on the correlation with the other parameters. This study adds the methodology of measuring corneal stiffness in vivo for ophthalmologists' reference in diagnosis.

Treatment time reduction for large thermal lesions by using a multiple 1D ultrasound phased array system.

To generate large thermal lesions in ultrasound thermal therapy, cooling intermissions are usually introduced during the treatment to prevent near-field heating, which leads to a long treatment time. A possible strategy to shorten the total treatment time is to eliminate the cooling intermissions. In this study, the two methods, power optimization and acoustic window enlargement, for reducing power accumulation in the near field are combined to investigate the feasibility of continuously heating a large target region (maximally 3.2 x 3.2 x 3.2 cm3). A multiple 1D ultrasound phased array system generates the foci to scan the target region. Simulations show that the target region can be successfully heated without cooling and no near-field heating occurs. Moreover, due to the fact that there is no cooling time during the heating sessions, the total treatment time is significantly reduced to only several minutes, compared to the existing several hours.

Novel approach to fuzzy-wavelet ECG signal analysis for a mobile device.

This paper describes a signal processing technique for ECG signal analysis based upon the combination of wavelet analysis and fuzzy c-means clustering. The signal analysis technique is implemented into a biomedical signal diagnostic unit that is the carry on device for the Wireless Nano-Bios Diagnostic System (WNBDS) developed at National Taiwan University. The WNBDS integrates mobile devices and remote data base servers to conduct online monitoring and remote healthcare applications. The signal analysis and diagnostic algorithms in this paper are implemented in an embedded mobile device to conduct mobile biomedical signal diagnostics. At this stage, the Electrocardiogram (ECG or EKG) is analyzed for patient health monitoring. The ECG signal processing is based on the wavelet analysis, and the diagnosis is based on fuzzy clustering. The embedded system is realized with the Windows CE operating system.

Research on three dimensional machining effects using atomic force microscope.

This research studies the use of scanning probe microscope as the tool to manufacture three dimensional nanoscale objects. We modified a commercial atomic force microscope (AFM) and replaced the original probe control system with a personal computer (PC) based controller. The modified system used the scanning probe in the AFM for the cutting tool and used the PC controller to control work piece. With the new controller, one could implement multiaxes motion control to perform trajectory planning and to test various cutting strategies. The experiments discovered that the debris can coalesce with the sample material and cause tremendous problem in the nanomachining process. This research thus proposed to make use of this material and developed a piling algorithm to not only cut but also pile up the debris in a favorable way for steric shaping. The experimental results showed that the proposed cutting and shaping algorithm can produce nano-objects as high as a few hundred nanometers. The probe tip typically wears down to around 500 microm diameter after the machining process, putting a limit on the machining resolution. The vertical resolution can achieve less than 10 nm without controlled environment.