Tzu-Chien Hsiao

Prediction of burn healing time using artificial neural networks and reflectance spectrometer

Burn depth assessment is important as early excision and grafting is the treatment of choice for deep dermal burn. Inaccurate assessment causes prolonged hospital stay, increased medical expenses and morbidity. Based on reflected burn spectra, we have developed an artificial neural network to predict the burn healing time. The purpose of our study is to develop a noninvasive objective method to predict burn-healing time. Burn less than 20% TBSA was included. Burn spectra taken on the third postburn day using reflectance spectrometer were analyzed by an artificial neural network system. 41 spectra were collected. With the newly developed method, the predictive accuracy of burns healed in less than 14 days was 96% and that in more than 14 days was 75%. Using reflectance spectrometer, we have developed an artificial neural network to determine the burn healing time with 86% overall predictive accuracy.

Partial least squares learning regression for backpropagation network

The relationship between the partial least squares (PLS) regression and the general delta rule algorithm is investigated. This PLS regression can be adopted as an efficient pre-learning method for backpropagation (BP) network. The PLS regression based BP network (PLSBP network) has better capacity during training phase. Aided by the statistical concept of the PLS regression, the cost function of this network is guaranteed to be an optimal minimum. The logistic map for network simulation is provided as an example.

The implementation of partial least squares with artificial neural network architecture

The widely used multivariate analysis method, partial least squares (PLS) regression is mapped to the general multilayer architecture of artificial neural networks. This architecture can be viewed as a parallel implementation of PLS method in the weight matrix of input-to-hidden layer. The nature of the PLS approach is comparable to the well-known backpropagation (BP) method, which also utilizes the input-output pair for error correction. This novel concept provides a way to view the statistical meaning of the extracted feature in BP method. Apart from the traditional views of principal component, which results from the autocorrelation of input patterns, this is the first time a different statistical description of the resultant weight matrix been proposed.

Rapid wheezing detection algorithm for real-time asthma diagnosis and personal health care

Pulmonary disease, such as asthma or chronic obstructive pulmonary disease (COPD), has been a major health concern throughtout the world. It would thus be necessary to develop an effective monitorning device and a real-time diagnosis algorithm for targeted populations, especially for children. Wheezing sound induced by asthma is a critical index for clinicians to make diagnosis. The traditional way to detect wheezing sound is to utilize the digital image process (DIP) method for tracking the specific wheezing pattern appeared in the short-time Fourier transform (STFT) spectral graph of respiratory sound. However, this method requires intensive computation and thus is difficult to implement for real-time diagnosis and low power consumption for personal health care system. In this study, we developed a new wheezing detection algorithm which is based on the estimation of correlation-coefficient of respiratory sound spectrum, called respiratory spectrum correlation-coefficient method (RSCM) in place of DIP step. Because of low memory quest of RSCM process, it can be installed easily in the microcontroller or PDA. We have implanted RSCM to a personal asthma daily care system based on both laptop and PDA. User can measure the respiratory sound by designed microphone input and real-time diagnose the occurrence of asthma. In the initial test, thirteen cases (six for wheezing and seven for normal) of respiratory sound were collected from the public domain websites. The result shows that the sensitivity and specificity for wheezing detection are 83% and 86%, respectively. This result assures the possibility to meet the demands of personal health care.

Fibreoptic fluorescence spectroscopy for monitoring fish freshness

In this study, a portable Y-type fibreoptic fluorescence spectroscopy measurement system was used to evaluate the freshness of eight cobias (Rachycentron canadum). The results showed that the ratio of fluorescent intensity, which F480 nm/Fexci+50 nm was belong with the range of collagen type I and type V characteristic spectra, was positive correlated to the frozen time by hours. It was a strong approach to be a potential index for differentiating the fish freshness during delivery process. Besides, the different pattern results of dorsum and abdomen were shown in this study. In further, fibreoptic fluorescence spectroscopy could be a way not only to measure and quantify the freshness of different fish body but also to verify the level of taste.

An implantable integrated SiGe FM transmitter for HRV biotelemetry

This paper presents an implantable integrated FM transmitter using 0.35 /spl mu/m SiGe process for biotelemetry. The architecture of FM transmitter is based on the Colpitts oscillator, which is followed by a buffer amplifier to isolate the influence of output on the resonant frequency of the oscillator. The chip area of the fabricated FM transmitter is only 0.26 mm/sup 2/. Even without an intentional antenna, the experimental results show that electrocardiograms (ECGs) of a rat using the FM transmitter can be successfully transmitted to a personal computer, where the heart rate variability (HRV) is analyzed.

Quantitative multivariate analysis with artificial neural networks

Quantitative interpretation of spectra can be achieved by using artificial neural networks with multi-layer architecture. Both back-propagation (BP) and radial basis function (RBF) are implemented and tested with raw absorption spectra and normalized spectra of glucose solutions in MATLAB. Simulation results showed that the partial least square (PLS) method can have a better performance with small number in the calibration set. However, with increasing size of data set, as in the cross validation method, RBF and BP have better performance. With optimal spreading factor, RBF can have the same degree of accuracy but significantly faster convergent speed comparing to BP. The normalization scheme can also significantly affect the performance of both RBF and BP.

Optical Characterization of a 1-D Nanostructure by Dark-Field Microscopy and Surface Plasmon Resonance to Determine Biomolecular Interactions

This paper presents a multifunctional imaging system that combines dark-field microscopy (DFM) with spectroscopy to image nanostructures and identify their optical properties from absorption spectra. The optical resolving power of this system is determined using a 1-D nanostructure with pitches of 120, 390, and 770 nm with four formats of optical disks. These pattern sizes are verified by atomic force microscopy (AFM) first. The results demonstrate that the resolving power of current system setup can down to 86 nm. The resultant DFM images appear to be slightly larger than the AFM images. A 50-nm-thick gold film was then deposited on top of these nanostructures, and their absorption spectra were obtained to elucidate its optical properties, enhanced by surface plasmon resonance. The immobilization of streptavidin on the surface of gold-coated nanostructure causes the absorption spectra to shift from 600 to 610 nm. A protein nanoarray with a dot size of 50 nm was also imaged by DFM, and can be implemented as a potential biochemical diagnostic system on an optical disk format. Specimens of adenocarcinoma cells and ovary cancer cells were also imaged using this DFM system, and the nuclei structure and some cellular organs can be recognized using a 100× objective oil lens.

Implementation of Smart Medical Home Gateway System for Chronic Patients

Because of the rapid aging population in Taiwan and the trend of fewer children, people are looking into technical solutions for continuous/intermittent monitoring of vital signs in the home setting environment and the interactions between family members. In this study, we have designed and implemented a home medical gateway system to connect the home-care side and the health informatics side. The home-care part provides five vital signs monitoring and on-line feedback message. Users are allowed to browse their records and read the received health information (e.g. physical checkup, health education, preventive inoculation...etc.) on the Flash based interface. This study also evaluated the practicability of the home gateway system. The number of interviewees is twenty. The analysis results show the positive user feedback of the system, and have high potential to promote the quality of patient’s life. An example case of obstructed sleep apnea (OSA) patient has been studied with this system. The result shows that the gateway system can help the OSA patient to monitor and improve their sleep quality.

Heart rate variability of internet gaming disorder addicts in emotional states

Internet gaming disorder (IGD) was regarded as behavioral addiction, and the common characteristics of internet addiction (IA) or IGD were craving, tolerance, withdrawal, mood modification, and salience. The fifth version of Diagnostic and Statistical Manual of Mental Disorders, addressed IGD, and also proposed 9 research criteria for IGD in 2013. Emotion was reported one character of IGD, and this study hypothesized that emotion was both a response and an influential factor on IGD. The self-report for emotion and questionnaires (psychological character), and electrocardiography (ECG) (physiological signal) of participants in emotional state were collected and analyzed in a non-invasive manner. An emotional induction experiment using online game film clips was conducted. Nineteen participants were grouped into IGD& high-risk internet addiction group, and 21 were in non-IGD & low-risk internet addiction group. The physiological signal (ECG) was used to study emotional responses of IGD. The sympathetic and parasympathetic activities were obtained from heart rate variability (HRV). The IGD addicts exhibited more positive emotion, and stronger sympathetic activity, but felt weaker physiological activity for online game than non-IGD addicts. Emotion can provide a short-term, quick response and dynamic changes of psychological and physiological information of IGD addicts, and that allowed researcher to develop further application, such as monitoring system, early alarm system, early IGD detection, and even prevention.