Browsing by Author "Liu, Chengyu"

Now showing 1 - 5 of 5

Biogenesis mechanisms of circular RNA can be categorized through feature extraction of a machine learning model
(2019-12-01)
Liu, Chengyu
;
Liu, Yu-Chen
;
Huang, Hsien-Da
;
Wang, Wei
;
生物資訊及系統生物研究所
Motivation: In recent years, multiple circular RNAs (circRNA) biogenesis mechanisms have been discovered. Although each reported mechanism has been experimentally verified in different circRNAs, no single biogenesis mechanism has been proposed that can universally explain the biogenesis of all tens of thousands of discovered circRNAs. Under the hypothesis that human circRNAs can be categorized according to different biogenesis mechanisms, we designed a contextual regression model trained to predict the formation of circular RNA from a random genomic locus on human genome, with potential biogenesis factors of circular RNA as the features of the training data. Results: After achieving high prediction accuracy, we found through the feature extraction technique that the examined human circRNAs can be categorized into seven subgroups, according to the presence of the following sequence features: RNA editing sites, simple repeat sequences, self-chains, RNA binding protein binding sites and CpG islands within the flanking regions of the circular RNA back-spliced junction sites. These results support all of the previously reported biogenesis mechanisms of circRNA and solidify the idea that multiple biogenesis mechanisms co-exist for different subset of human circRNAs. Furthermore, we uncover a potential new links between circRNA biogenesis and flanking CpG island. We have also identified RNA binding proteins putatively correlated with circRNA biogenesis.
Design of Novel Field Programmable Gate Array-Based Hearing Aid
(2019-01-01)
Lin, Bor-Shing
;
Yang, Po-Yu
;
Liu, Ching-Feng
;
Huang, Yi-Chia
;
Liu, Chengyu
Hearing loss is one of the most common chronic diseases. For people with hearing loss, communicating with other people, particularly in an environment with considerable background noise, is difficult. Recently, several hearing aids have been developed to improve speech comprehension in a noisy environment. The use of an adaptive beamformer is one of the alternative methods for improving speech intelligibility. However, the adaptive beamformer requires the location of the desired speaker to estimate the time differences of arrival (TDOAs) of speech sources to numerous spatially separated sensors in acoustics. In general, the technique of steered response power source localization was used to estimate the TDOA; however, this technique was easily affected by environmental reverberation. To overcome the aforementioned concern, a novel hearing aid is proposed in this paper. By using an image processing technology, the location of the desired speaker could be manually selected to provide precise information on the TDOA. Moreover, adaptive signal enhancements were implemented in a field-programmable gate array to enhance the speech of interest in real time. The experimental results indicate that the proposed system could improve speech intelligibility in various noisy environments. Therefore, the proposed system may be employed to improve the daily lives of people with hearing the loss in the future.
Development of Novel Hearing Aids by Using Image Recognition Technology
(2019-05-01)
Lin, Bor-Shing
;
Liu, Ching-Feng
;
Cheng, Chih-Jen
;
Wang, Jhi-Joung
;
Liu, Chengyu
Speech is easily affected by different background noise in real environment to reduce the speech intelligibility, in particular, for hearing impaired listeners. In order to improve the above issue, several hearing aids have been developed to enhance the speech signal in noisy environment. Most of current hearing aids were designed to enhance the component of speech and suppress the component of noise. However, it is difficult to separate other speech sources. Adaptive signal enhancement with the beamforming technique might improve the above issue. However, how to distinguish the location of the desired speaker effectively is still a difficult challenge for adaptive beamforming method. A novel concept of hearing aid was proposed in this study. Different from the beamforming-based hearing aids, which use the cross-correlation-coefficient method to estimate time difference of arrival (TDOA), an image recognition technology was used to estimate the location of the desired speaker to obtain the more precise TDOA. An adaptive signal enhancement was also used to enhance the noisy speech sound. From the experimental results, the proposed system could provide a smaller absolute error of TDOA less than 1.25 x 10(-4) ms, and a clear speech sound from the target speaker who the user wants to listen to.
Effects of Inferior Myocardial Infarction Sizes and Sites on Simulated Electrocardiograms Based on a Torso-Heart Model
(2019-01-01)
Cai, Zhipeng
;
Li, Jianqing
;
Luo, Kan
;
Lin, Bor-Shyh
;
Wang, Zhigang
Myocardial infarction (MI) is one of the major causes of death. Thus, understanding the underlying mechanisms of MI and its clinical features, especially its relationship with common electrocardiography measurements, is important. Heart modeling provides a possible method to simulate certain heart conditions. In this paper, a refined MI torso-heart model was proposed to explore the effect of inferior MI on simulated electrocardiograms. This model simulated the electrical activity of a normal heart and an inferior MI heart at three sites (basal, middle, and apical) with three MI sizes (small, moderate, and large), aiming to investigate the effect of inferior MI on the simulated electrocardiograms. Simulated body surface potential maps were compared with clinical body surface potential maps to verify the effciency of the model. A new measure, namely, the normalized variation coeffcient, was proposed for result evaluation. The results showed that the augmented unipolar left lower limb lead was the best choice for inferior MI diagnoses and it showed the most obvious electrocardiography signal differences between normal and inferior MI hearts. The simulation results corresponded well with commonly used clinical diagnostic criteria. Thus, the proposed refined MI torso-heart model provides a finite element simulation method for quantifying the effects of inferior MI on a torso-heart model-based electrocardiogram and has good potential for use in optimizing electrocardiogram detection.
Noise Rejection for Wearable ECGs Using Modeified Frequency Slice Wavelet Transform and Convolutional Neural Networks
(2019-01-01)
Zhao, Zhongyao
;
Liu, Chengyu
;
Li, Yaowei
;
Li, Yixuan
;
Wang, Jingyu
Progress in wearable techniques makes the long-term daily electrocardiogram (ECG) monitoring possible. However, the long-term wearable ECGs can be significantly contaminated by various noises, which affect the detection and diagnosis of cardiovascular diseases (CVDs). The situation becomes more serious for wearable ECG screening, where the data are huge, and doctors have no way to visually check the signal quality episode-by-episode. Therefore, automatic and accurate noise rejection for the wearable big-data ECGs is craving. This paper addressed this issue and proposed a noise rejection method for wearable ECGs based on the combination of modified frequency slice wavelet transform (MFSWT) and convolutional neural network (CNN). Wearable ECGs were recorded using the newly developed 12-lead Lenovo smart ECG vest with a sample rate of 500 Hz and a resolution of 16 bit. One thousand 10-s ECG segments were picked up and were manually labeled into three quality types: clinically useful segments with good signal quality (type A), clinically useful segments with poor signal quality (type B), and clinically useless segments (pure noises, type C). Each of the 1,000 10-s ECG segments were transformed into a 2-D time-frequency (T-F) image using the MFSWT, with a pixel size of 200x50. Then, the 2-D grayscale images from MFSWT were fed into a 13-layer CNN model for training the classification models. Results from the standard 5-folder cross-validation showed that the proposed combination method of MFSWT and CNN achieved a highest classification accuracy of 86.3%, which was higher than the comparable methods from continuous wavelet transform (CWT) and artificial neural networks (ANN). The combination of MFSWT and CNN also had a good calculation efficiency. This paper indicated that the combination of MFSWT and CNN is a potential method for automatic identification of noisy segments from wearable ECG recordings.