Lightweight Deep Neural Network Embedded with Stochastic Variational Inference Loss Function for Fast Detection of Human Postures.

Fusing object detection techniques and stochastic variational inference, we proposed a new scheme for lightweight neural network models, which could simultaneously reduce model sizes and raise the inference speed. This technique was then applied in fast human posture identification. The integer-arithmetic-only algorithm and the feature pyramid network were adopted to reduce the computational complexity in training and to capture features of small objects, respectively. Features of sequential human motion frames (i.e., the centroid coordinates of bounding boxes) were extracted by the self-attention mechanism. With the techniques of Bayesian neural network and stochastic variational inference, human postures could be promptly classified by fast resolving of the Gaussian mixture model for human posture classification. The model took instant centroid features as inputs and indicated possible human postures in the probabilistic maps. Our model had better overall performance than the baseline model ResNet in mean average precision (32.5 vs. 34.6), inference speed (27 vs. 48 milliseconds), and model size (46.2 vs. 227.8 MB). The model could also alert a suspected human falling event about 0.66 s in advance.

How to Implement Automotive Fault Diagnosis Using Artificial Intelligence Scheme.

The necessity of vehicle fault detection and diagnosis (VFDD) is one of the main goals and demands of the Internet of Vehicles (IoV) in autonomous applications. This paper integrates various machine learning algorithms, which are applied to the failure prediction and warning of various types of vehicles, such as the vehicle transmission system, abnormal engine operation, and tire condition prediction. This paper first discusses the three main AI algorithms, such as supervised learning, unsupervised learning, and reinforcement learning, and compares the advantages and disadvantages of each algorithm in the application of system prediction. In the second part, we summarize which artificial intelligence algorithm architectures are suitable for each system failure condition. According to the fault status of different vehicles, it is necessary to carry out the evaluation of the digital filtering process. At the same time, it is necessary to preconstruct its model analysis and adjust the parameter attributes, types, and number of samples of various vehicle prediction models according to the analysis results, followed by optimization to obtain various vehicle models. Finally, through a cross-comparison and sorting, the artificial intelligence failure prediction models can be obtained, which can correspond to the failure status of a certain car model and a certain system, thereby realizing a most appropriate AI model for a specific application.

Deep Learning with LPC and Wavelet Algorithms for Driving Fault Diagnosis.

Vehicle fault detection and diagnosis (VFDD) along with predictive maintenance (PdM) are indispensable for early diagnosis in order to prevent severe accidents due to mechanical malfunction in urban environments. This paper proposes an early voiceprint driving fault identification system using machine learning algorithms for classification. Previous studies have examined driving fault identification, but less attention has focused on using voiceprint features to locate corresponding faults. This research uses 43 different common vehicle mechanical malfunction condition voiceprint signals to construct the dataset. These datasets were filtered by linear predictive coefficient (LPC) and wavelet transform(WT). After the original voiceprint fault sounds were filtered and obtained the main fault characteristics, the deep neural network (DNN), convolutional neural network (CNN), and long short-term memory (LSTM) architectures are used for identification. The experimental results show that the accuracy of the CNN algorithm is the best for the LPC dataset. In addition, for the wavelet dataset, DNN has the best performance in terms of identification performance and training time. After cross-comparison of experimental results, the wavelet algorithm combined with DNN can improve the identification accuracy by up to 16.57% compared with other deep learning algorithms and reduce the model training time by up to 21.5% compared with other algorithms. Realizing the cross-comparison of recognition results through various machine learning methods, it is possible for the vehicle to proactively remind the driver of the real-time potential hazard of vehicle machinery failure.

IC-Based Rectification Circuit Techniques for Biomedical Energy-Harvesting Applications.

Energy harvesting can be achieved through many different mechanisms. Such technology has been drawing researchers' attention to its practical applications for a decade, as it can be widely applied to countless scenarios. It steals the show in the modern development of the biomedical electronics, especially implantable applications, as it allows the patients to move freely without restriction. To prolong lifetime of the battery inside/outside a patient's body, the electrical conversion efficiency of the electronic implant is of primary importance in energy harvesting. The conversion can be achieved by a so-called miniaturized rectification circuit (also known as "rectifier"). This study aims to compare different state-of-the-art techniques focusing on the conversion efficiency of the rectification. Particular emphasis is put on semiconductor-based circuits capable of being integrated with tiny chips on the implants.

Exploiting deep neural network and long short-term memory method-ologies in bioacoustic classification of LPC-based features.

The research describes the recognition and classification of the acoustic characteristics of amphibians using deep learning of deep neural network (DNN) and long short-term memory (LSTM) for biological applications. First, original data is collected from 32 species of frogs and 3 species of toads commonly found in Taiwan. Secondly, two digital filtering algorithms, linear predictive coding (LPC) and Mel-frequency cepstral coefficient (MFCC), are respectively used to collect amphibian bioacoustic features and construct the datasets. In addition, principal component analysis (PCA) algorithm is applied to achieve dimensional reduction of the training model datasets. Next, the classification of amphibian bioacoustic features is accomplished through the use of DNN and LSTM. The Pytorch platform with a GPU processor (NVIDIA GeForce GTX 1050 Ti) realizes the calculation and recognition of the acoustic feature classification results. Based on above-mentioned two algorithms, the sound feature datasets are classified and effectively summarized in several classification result tables and graphs for presentation. The results of the classification experiment of the different features of bioacoustics are verified and discussed in detail. This research seeks to extract the optimal combination of the best recognition and classification algorithms in all experimental processes.

Ultrasound Detection of Liver Fibrosis in Individuals with Hepatic Steatosis Using the Homodyned K Distribution.

Acoustic structure quantification (ASQ) based on the analysis of ultrasound backscattered statistics has been reported to detect liver fibrosis without significant hepatic steatosis. This study proposed using ultrasound parametric imaging based on the parameter α of the homodyned K (HK) distribution for staging liver fibrosis in patients with significant hepatic steatosis. Raw ultrasound image data were acquired from patients (n = 237) to construct B-mode and HK α parametric images, which were compared with the focal disturbance (FD) ratio obtained from ASQ on the basis of histologic evidence (METAVIR fibrosis score and hepatic steatosis severity). The data were divided into group I (n = 173; normal to mild hepatic steatosis) and group II (n = 64; with moderate to severe hepatic steatosis) for statistical analysis through one-way analysis of variance and receiver operating characteristic (ROC) curve analysis. The results showed that the HK α parameter monotonically decreased as the liver fibrosis stage increased (p < .05); concurrently, the FD ratio increased (p < .05). For group I, the areas under the ROC (AUROCs) obtained using the FD ratio and the α parameter (AUROCFD and AUROCα) were, respectively, 0.56 and 0.55, 0.68 and 0.68, 0.64 and 0.64 and 0.62 and 0.62 for diagnosing liver fibrosis ≥F1, ≥F2, ≥F3 and ≥F4. The values of AUROCFD and AUROCα for group II were, respectively, 0.88 and 0.91, 0.81 and 0.81, 0.77 and 0.76 and 0.78 and 0.73 for diagnosing liver fibrosis ≥F1, ≥F2, ≥F3 and ≥F4. As opposed to previous studies, ASQ was found to fail in characterizing liver fibrosis in group I; however, it was workable for identifying liver fibrosis in patients with significant hepatic steatosis (group II). Compared with ASQ, HK imaging provided improved diagnostic performance in the early detection of liver fibrosis coexisting with moderate to severe hepatic steatosis. Ultrasound HK imaging is recommended as a strategy to evaluate early fibrosis risk in patients with significant hepatic steatosis.

A Bootstrapped Comparator-Switched Active Rectifying Circuit for Wirelessly Powered Integrated Miniaturized Energy Sensing Systems.

Human life expectancy has gradually increased in part through rapid advances in technology, including the development and use of wearable and implantable biomedical electronic devices and sensing monitors. A new architecture is proposed in this paper to replace the traditional diode circuit implementation in wireless power supply systems applied to the above-mentioned devices and monitors. By achieving near-ideal power transistor switching and leveraging the characteristics of conventional diodes to prevent reverse current, the proposed approach greatly improves the performance of the energy harvester in power conversion. The MOS harvester used in the uninterrupted permanent wireless near-field power supply described here for use in biomedical systems was designed and verified using the Taiwan Semiconductor Manufacturing Company (TSMC) standard 180-nm process, achieving performance results of Voltage conversion efficiency (VCE) = 73.55-95.12% and Power conversion efficiency (PCE) = 80.36-90.08% with the output load (0.1-1 kΩ) under 3.3 V ac input with an overall area of 1.189 mm2. These results are expected to create an important technical niche for new "green-energy" miniaturized energy sensing systems including cutting edge wirelessly powered biomedical electronics applications.

Optically Guided Epidural Needle Placement Using 405-nm Wavelength for Accurate Puncture.

Several approaches of locating the epidural space have been proposed. However, loss of Resistance method (LOR) remains the most common method for epidural anesthesia. Different optical signals were received from the ligamentum flavum and the epidural space allows operator to pinpoint position of the needle and determine whether the needle tip has entered the epidural space. Optical signals throughout the penetration process was recorded and position of needle tip was confirmed with a C-arm fluoroscopy. 60 lumbar punctures were performed in 20 vivo porcine models, and success rate of locating the epidural space with the optical auxiliary is calculated statistically. The data are expressed in mean ± SD. During all the lumber puncture processes, the strength of optical signals received decreased significantly while the needle tip penetrates the ligamentum flavum and entered the epidural space. The strength of optical signal received when needle tip was in the ligamentum flavum was 1.38 ± 0.57. The signal strength at epidural space was 0.46 ± 0.35. Strength of signal decreased by 67% when entered epidural space, and there is no significant differences in decrease of strength from data obtained from thevertebrae (lumbar segments)L2-L3, L3-L4, and L4-L5. Finally, we calculated with assistance of the proposed optical auxiliary, the success rate for guiding the needle tip to the epidural space using was as high as 87%. It is evidently believed that the optical auxiliary equipped is visualized to assist operators inserting needle accurately and efficiently into epidural space during epidural anesthesia operation.

Double Assurance of Epidural Space Detection Using Fiberoptics-Based Needle Design and Autofluorescence Technologies for Epidural Blockade in Painless Labor.

Purpose: Technology of reflectance spectroscopy incorporated with auto-fluorescence spectroscopy were employed to increase the safety of epidural placement in regional anesthesia which is generally used for surgery, epidural anesthesia, post-operative pain control and painless childbirth. Method: Ex vivo study of auto-fluorescence spectroscopy was performed for the para-vertebral tissues contained fat, interspinous ligament, supraspinous ligament and ligamentumflavum by multimode microplate reader at wavelength 405 nm for the purpose of tissue differentiation. A specially designed optic-fiber-embedded needle was employed to incorporate with both reflectance and autofluorescence spectroscopies in order to probe the epidural space as double assurance demands. In vivo study was carried out in a Chinese native swine weighted about 30 kg under intubated general anesthesia with ventilation support. The reflective (405 nm) and autofluorescence signals (λ and λ*) were recorded at 5 different sites by an oscilloscope during the needle puncture procedure from skin to epidural space in the back of the swine. Results: Study of either autofluorescence spectroscopy for tissue samples or ex vivo needle puncture in porcine trunk tissues indicates that ligmentumflavum has at least 10-fold higher fluorescence intensity than the other tissues. In the in vivo study, ligamentumflavum shows a double-peak character for both reflectance and autofluorescence signals. The epidural space is located right after the drop from the double-peak. Both peaks of reflectance and fluorescence are coincident which ensures that the epidural space is correctly detected. Conclusions: The fiber-optical technologies of double-assurance demands for tissue discrimination during epidural needle puncture can not only provide an objective visual information in a real-time fashion but also it can help the operator to achieve much higher success rate in this anesthesia procedure.

Development of a Flexible Non-Metal Electrode for Cell Stimulation and Recording.

This study presents a method of producing flexible electrodes for potentially simultaneously stimulating and measuring cellular signals in retinal cells. Currently, most multi-electrode applications rely primarily on etching, but the metals involved have a certain degree of brittleness, leaving them prone to cracking under prolonged pressure. This study proposes using silver chloride ink as a conductive metal, and polydimethysiloxane (PDMS) as the substrate to provide electrodes with an increased degree of flexibility to allow them to bend. This structure is divided into the electrode layer made of PDMS and silver chloride ink, and a PDMS film coating layer. PDMS can be mixed in different proportions to modify the degree of rigidity. The proposed method involved three steps. The first segment entailed the manufacturing of the electrode, using silver chloride ink as the conductive material, and using computer software to define the electrode size and micro-engraving mechanisms to produce the electrode pattern. The resulting uniform PDMS pattern was then baked onto the model, and the flow channel was filled with the conductive material before air drying to produce the required electrode. In the second stage, we tested the electrode, using an impedance analyzer to measure electrode cyclic voltammetry and impedance. In the third phase, mechanical and biocompatibility tests were conducted to determine electrode properties. This study aims to produce a flexible, non-metallic sensing electrode which fits snugly for use in a range of measurement applications.

The High-efficiency LED Driver for Visible Light Communication Applications.

This paper presents a LED driver for VLC. The main purpose is to solve the low data rate problem used to be in switching type LED driver. The GaN power device is proposed to replace the traditional silicon power device of switching LED driver for the purpose of increasing switching frequency of converter, thereby increasing the bandwidth of data transmission. To achieve high efficiency, the diode-connected GaN power transistor is utilized to replace the traditional ultrafast recovery diode used to be in switching type LED driver. This work has been experimentally evaluated on 350-mA output current. The results demonstrate that it supports the data of PWM dimming level encoded in the PPM scheme for VLC application. The experimental results also show that system's efficiency of 80.8% can be achieved at 1-Mb/s data rate.

Predicting postoperative vomiting among orthopedic patients receiving patient-controlled epidural analgesia using SVM and LR.

Patient-controlled epidural analgesia (PCEA) has been applied to reduce postoperative pain in orthopedic surgical patients. Unfortunately, PCEA is occasionally accompanied by nausea and vomiting. The logistic regression (LR) model is widely used to predict vomiting, and recently support vector machines (SVM), a supervised machine learning method, has been used for classification and prediction. Unlike our previous work which compared Artificial Neural Networks (ANNs) with LR, this study uses a SVM-based predictive model to identify patients with high risk of vomiting during PCEA and comparing results with those derived from the LR-based model. From January to March 2007, data from 195 patients undergoing PCEA following orthopedic surgery were applied to develop two predictive models. 75% of the data were randomly selected for training, while the remainder was used for testing to validate predictive performance. The area under curve (AUC) was measured using the Receiver Operating Characteristic curve (ROC). The area under ROC curves of LR and SVM models were 0.734 and 0.929, respectively. A computer-based predictive model can be used to identify those who are at high risk for vomiting after PCEA, allowing for patient-specific therapeutic intervention or the use of alternative analgesic methods.

A programmable high-voltage compliance neural stimulator for deep brain stimulation in vivo.

Deep brain stimulation (DBS) is one of the most effective therapies for movement and other disorders. The DBS neurosurgical procedure involves the implantation of a DBS device and a battery-operated neurotransmitter, which delivers electrical impulses to treatment targets through implanted electrodes. The DBS modulates the neuronal activities in the brain nucleus for improving physiological responses as long as an electric discharge above the stimulation threshold can be achieved. In an effort to improve the performance of an implanted DBS device, the device size, implementation cost, and power efficiency are among the most important DBS device design aspects. This study aims to present preliminary research results of an efficient stimulator, with emphasis on conversion efficiency. The prototype stimulator features high-voltage compliance, implemented with only a standard semiconductor process, without the use of extra masks in the foundry through our proposed circuit structure. The results of animal experiments, including evaluation of evoked responses induced by thalamic electrical stimuli with our fabricated chip, were shown to demonstrate the proof of concept of our design.

Advances in miniaturized instruments for genomics.

In recent years, a lot of demonstrations of the miniaturized instruments were reported for genomic applications. They provided the advantages of miniaturization, automation, sensitivity, and specificity for the development of point-of-care diagnostics. The aim of this paper is to report on recent developments on miniaturized instruments for genomic applications. Based on the mature development of microfabrication, microfluidic systems have been demonstrated for various genomic detections. Since one of the objectives of miniaturized instruments is for the development of point-of-care device, impedimetric detection is found to be a promising technique for this purpose. An in-depth discussion of the impedimetric circuits and systems will be included to provide total consideration of the miniaturized instruments and their potential application towards real-time portable imaging in the "-omics" era. The current excellent demonstrations suggest a solid foundation for the development of practical and widespread point-of-care genomic diagnostic devices.

Predicting postoperative vomiting for orthopedic patients receiving patient-controlled epidural analgesia with the application of an artificial neural network.

Patient-controlled epidural analgesia (PCEA) was used in many patients receiving orthopedic surgery to reduce postoperative pain but is accompanied with certain incidence of vomiting. Predictions of the vomiting event, however, were addressed by only a few authors using logistic regression (LR) models. Artificial neural networks (ANN) are pattern-recognition tools that can be used to detect complex patterns within data sets. The purpose of this study was to develop the ANN based predictive model to identify patients with high risk of vomiting during PCEA used. From January to March 2007, the PCEA records of 195 patients receiving PCEA after orthopedic surgery were used to develop the two predicting models. The ANN model had a largest area under curve (AUC) in receiver operating characteristic (ROC) curve. The areas under ROC curves of ANN and LR models were 0.900 and 0.761, respectively. The computer-based predictive model should be useful in increasing vigilance in those patients most at risk for vomiting while PCEA is used, allowing for patient-specific therapeutic intervention, or even in suggesting the use of alternative methods of analgesia.

Portable optical epidural needle-a CMOS-based system solution and its circuit design.

Epidural anesthesia is a common anesthesia method yet up to 10% of procedures fail to provide adequate analgesia. This is usually due to misinterpreting the tactile information derived from the advancing needle through the complex tissue planes. Incorrect placement also can cause dural puncture and neural injury. We developed an optic system capable of reliably identifying tissue planes surrounding the epidural space. However the new technology was too large and cumbersome for practical clinical use. We present a miniaturized version of our optic system using chip technology (first generation CMOS-based system) for logic functions. The new system was connected to an alarm that was triggered once the optic properties of the epidural were identified. The aims of this study were to test our miniaturized system in a porcine model and describe the technology to build this new clinical tool. Our system was tested in a porcine model and identified the epidural space in the lumbar, low and high thoracic regions of the spine. The new technology identified the epidural space in all but 1 of 46 attempts. Experimental results from our fabricated integrated circuit and animal study show the new tool has future clinical potential.

Investigation of temporal vascular effects induced by focused ultrasound treatment with speckle-variance optical coherence tomography.

Focused ultrasound (FUS) can be used to locally and temporally enhance vascular permeability, improving the efficiency of drug delivery from the blood vessels into the surrounding tissue. However, it is difficult to evaluate in real time the effect induced by FUS and to noninvasively observe the permeability enhancement. In this study, speckle-variance optical coherence tomography (SVOCT) was implemented for the investigation of temporal effects on vessels induced by FUS treatment. With OCT scanning, the dynamic change in vessels during FUS exposure can be observed and studied. Moreover, the vascular effects induced by FUS treatment with and without the presence of microbubbles were investigated and quantitatively compared. Additionally, 2D and 3D speckle-variance images were used for quantitative observation of blood leakage from vessels due to the permeability enhancement caused by FUS, which could be an indicator that can be used to determine the influence of FUS power exposure. In conclusion, SVOCT can be a useful tool for monitoring FUS treatment in real time, facilitating the dynamic observation of temporal effects and helping to determine the optimal FUS power.

Investigation of an intelligent system for fiber optic-based epidural anesthesia.

Even though there have been many approaches to assist the anesthesiologists in performing regional anesthesia, none of the prior arts may be said as an unrestricted technique. The lack of a design that is with sufficient sensitivity to the targets of interest and automatic indication of needle placement makes it difficult to all-round implementation of field usage of objectiveness. In addition, light-weight easy-to-use realization is the key point of portability. This paper reports on an intelligent system of epidural space identification using optical technique, with particular emphasis on efficiency-enhanced aspects. Statistical algorithms, implemented in a dedicated field-programmable hardware platform along with an on-platform application-specific integrated chip, used to advance real-time self decision making in needle advancement are discussed together with the feedback results. Clinicians' viewpoint of improving the correct rate of our technique is explained in detail. Our study demonstrates not only that the improved system is able to behave as if it is a skillful anesthesiologist but also it has potential to bring promising assist into clinical use under varied conditions and small amount of sample, provided that several concerns are addressed.