Abstract: Alzheimer’s disease (AD) is one of the deadliest neurodegenerative diseases ailing the elderly population all over the world. Many researchers are using deep learning (DL) techniques to learn highly complicated patterns from MRI scans for the detection of AD. It is also found that an ensemble of predictions from multiple models gives better performance as compared to that of a single model. Two major bottlenecks for developing ensemble of DL models are their high computational complexity and requirement of large sample size for better generalization. In this work, we deal with the aforementioned bottlenecks and propose a computationally efficient, DL-architecture agnostic, ensemble of deep neural networks named ‘Deep Transfer Ensemble (DTE)’ trained using transfer learning for the classification of AD. The proposed ensemble leverages the diversity introduced by many different locally optimum solutions reached by individual networks through the randomization of hyper-parameters. The proposed ensemble model also introduces further diverse predictions by exploiting complementary feature views. We also test the model vigorously by analyzing its performance on a large and a small dataset downloaded from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) archive. The DTE utilizes the advantages of random search, transfer learning, and snapshot ensembles in a single ensemble to produce better generalization performance. DTE achieves an accuracy of 99.05% and 85.27% on two independent splits of the large dataset for cognitively normal (NC) vs AD classification task. For the task of mild cognitive impairment (MCI) vs AD classification, DTE achieves 98.71% and 83.11% respectively on the two independent splits. DTE also performed reasonably well on a small dataset consisting of only 50 samples per class. It achieved a maximum accuracy of 85% for NC vs AD on the small dataset. DTE outperformed snapshot ensembles along with several other existing deep models from similar kind of previous works by other researchers.

Abstract: Recently, as one of the most promising biometric traits, the vein has attracted the attention of both academia and industry because of its living body identification and the convenience of the acquisition process. State-of-the-art techniques can provide relatively good performance, yet they are limited to specific light sources. Besides, it still has poor adaptability to multispectral images. Despite the great success achieved by convolutional neural networks (CNNs) in various image understanding tasks, they often require large training samples and high computation that are infeasible for palm-vein identification. To address this limitation, this work proposes a palm-vein identification system based on lightweight CNN and adaptive multi-spectral method with explainable AI. The principal component analysis on symmetric discrete wavelet transform (SMDWT-PCA) technique for vein images augmentation method is adopted to solve the problem of insufficient data and multispectral adaptability. The depth separable convolution (DSC) has been applied to reduce the number of model parameters in this work. To ensure that the experimental result demonstrates accurately and robustly, a multispectral palm image of the public dataset (CASIA), is also used to assess the performance of the proposed method. As result, the palm-vein identification system can provide superior performance to that of the former related approaches for different spectrums.

Abstract: Background and purpose The left anterior descending coronary artery (LAD) and diagonal branches (DBs) are blurred on computed tomography (CT). We aimed to define the LAD region (LADR) with adequate inclusion of the LAD and DBs and contouring consistency.

Methods and materials
The LADR was defined using coronary CT angiograms. The inclusion ratio was used to assess the LAD and DBs inclusion by the LADR. Four radiation oncologists delineated the LAD and LADR, using contrast-enhanced CT of 15 patients undergoing left breast radiotherapy. The Sørensen–Dice similarity index (DSI), Jaccard similarity index (JSI), and Hausdorff distance (HD) were calculated to assess similarity. The mean dose (Dmean) and maximum dose (Dmax) to the LAD and LADR were calculated to compare consistency. Correlations were evaluated using Pearson’s correlation coefficient.

Results
The inclusion ratio of the LAD by the LADR was 96%. The mean DSI, JSI, and HD values were respectively 27.9%, 16.7%, and 0.42 mm for the LAD, and 83.1%, 73.0%, and 0.18 mm for the LADR. The Dmean between the LAD and LADR were strongly correlated (r = 0.93).

Conclusion
Delineation of the LADR significantly improved contouring similarity and consistency for dose reporting. This could optimize dose estimation for breast radiotherapy.

Abstract: Digital dental reconstruction can be a more efficient and effective mechanism for artificial crown construction and period inspection. However, optical methods cannot reconstruct those portions under gums, and X-ray-based methods have high radiation to limit their applied frequency. Optical coherence tomography (OCT) can harmlessly penetrate gums using low-coherence infrared rays, and thus, this work designs an OCT-based framework for dental reconstruction using optical rectification, fast Fourier transform, volumetric boundary detection, and Poisson surface reconstruction to overcome noisy imaging. Additionally, in order to operate in a patient’s mouth, the caliber of the injector is small along with its short penetration depth and effective operation range, and thus, reconstruction requires multiple scans from various directions along with proper alignment. However, flat regions, such as the mesial side of front teeth, may not have enough features for alignment. As a result, we design a scanning order for different types of teeth starting from an area of abundant features for easier alignment while using gyros to track scanned postures for better initial orientations. It is important to provide immediate feedback for each scan, and thus, we accelerate the entire signal processing, boundary detection, and point-cloud alignment using Graphics Processing Units (GPUs) while streamlining the data transfer and GPU computations. Finally, our framework can successfully reconstruct three isolated teeth and a side of one living tooth with comparable precisions against the state-of-art method. Moreover, a user study also verifies the effectiveness of our interactive feedback for efficient and fast clinic scanning.

Abstract: Periodontal diagnosis requires discovery of the relations among teeth, gingiva (i.e., gums), and alveolar bones, but alveolar bones are inside gingiva and not visible for inspection. Traditional probe examination causes pain, and X-ray based examination is not suited for frequent inspection. This work develops an automatic non-invasive periodontal inspection framework based on gum penetrative Optical Coherence Tomography (OCT), which can be frequently applied without high radiation. We sum up interference responses of all penetration depths for all shooting directions respectively to form the shooting amplitude projection. Because the reaching interference strength decays exponentially with tissues’ penetration depth, this projection mainly reveals the responses of the top most gingiva or teeth. Since gingiva and teeth have different air-tissue responses, the gumline, revealing itself as an obvious boundary between teeth and gingiva, is the basis line for periodontal inspection. Our system can also automatically identify regions of gingiva, teeth, and alveolar bones from slices of the cross-sectional volume. Although deep networks can successfully and possibly segment noisy maps, reducing the number of manually labeled maps for training is critical for our framework. In order to enhance the effectiveness and efficiency of training and classification, we adjust Snake segmentation to consider neighboring slices in order to locate those regions possibly containing gingiva-teeth and gingiva–alveolar boundaries. Additionally, we also adapt a truncated direct logarithm based on the Snake-segmented region for intensity quantization to emphasize these boundaries for easier identification. Later, the alveolar-gingiva boundary point directly under the gumline is the desired alveolar sample, and we can measure the distance between the gumline and alveolar line for visualization and direct periodontal inspection. At the end, we experimentally verify our choice in intensity quantization and boundary identification against several other algorithms while applying the framework to locate gumline and alveolar line in vivo data successfully.

Abstract: Digital dental reconstruction can be a more efficient and effective mechanism for artificial crown construction and period inspection. However, optical methods cannot reconstruct those portions under gums, and X-ray-based methods have high radiation to limit their applied frequency. Optical coherence tomography (OCT) can harmlessly penetrate gums using low-coherence infrared rays, and thus, this work designs an OCT-based framework for dental reconstruction using optical rectification, fast Fourier transform, volumetric boundary detection, and Poisson surface reconstruction to overcome noisy imaging. Additionally, in order to operate in a patient’s mouth, the caliber of the injector is small along with its short penetration depth and effective operation range, and thus, reconstruction requires multiple scans from various directions along with proper alignment. However, flat regions, such as the mesial side of front teeth, may not have enough features for alignment. As a result, we design a scanning order for different types of teeth starting from an area of abundant features for easier alignment while using gyros to track scanned postures for better initial orientations. It is important to provide immediate feedback for each scan, and thus, we accelerate the entire signal processing, boundary detection, and point-cloud alignment using Graphics Processing Units (GPUs) while streamlining the data transfer and GPU computations. Finally, our framework can successfully reconstruct three isolated teeth and a side of one living tooth with comparable precisions against the state-of-art method. Moreover, a user study also verifies the effectiveness of our interactive feedback for efficient and fast clinic scanning.

Abstract: Periodontal diagnosis requires discovery of the relations among teeth, gingiva (i.e., gums), and alveolar bones, but alveolar bones are inside gingiva and not visible for inspection. Traditional probe examination causes pain, and X-ray based examination is not suited for frequent inspection. This work develops an automatic non-invasive periodontal inspection framework based on gum penetrative Optical Coherence Tomography (OCT), which can be frequently applied without high radiation. We sum up interference responses of all penetration depths for all shooting directions respectively to form the shooting amplitude projection. Because the reaching interference strength decays exponentially with tissues’ penetration depth, this projection mainly reveals the responses of the top most gingiva or teeth. Since gingiva and teeth have different air-tissue responses, the gumline, revealing itself as an obvious boundary between teeth and gingiva, is the basis line for periodontal inspection. Our system can also automatically identify regions of gingiva, teeth, and alveolar bones from slices of the cross-sectional volume. Although deep networks can successfully and possibly segment noisy maps, reducing the number of manually labeled maps for training is critical for our framework. In order to enhance the effectiveness and efficiency of training and classification, we adjust Snake segmentation to consider neighboring slices in order to locate those regions possibly containing gingiva-teeth and gingiva–alveolar boundaries. Additionally, we also adapt a truncated direct logarithm based on the Snake-segmented region for intensity quantization to emphasize these boundaries for easier identification. Later, the alveolar-gingiva boundary point directly under the gumline is the desired alveolar sample, and we can measure the distance between the gumline and alveolar line for visualization and direct periodontal inspection. At the end, we experimentally verify our choice in intensity quantization and boundary identification against several other algorithms while applying the framework to locate gumline and alveolar line in vivo data successfully.

Abstract: Our current project involves the development of a wristband-mounted sensor that is meant to function as an alcohol use monitoring system. This paper focuses on the degree to which physical activity influences ethanol concentrations in the vapor secreted from the skin through collecting data from seven recruited participants when they conducting one designated activity, which could presumably affect the accuracy of detection results. We proposes a preliminary design of building a personal alcohol tracking system that can improve the reliability and affordability of current transdermal ethanol tracking devices to accommodate potential interferences presented in daily life and be intuitive to be used to raise the awareness of alcohol use.

Abstract: In this study, we propose a new model aiming to enhance the quality of entity linking by exploiting highly relevant hyperlinks in knowledge base for entity disambiguation. We find that most existing studies do not filter the corresponding hyperlinks for each entity, where using the irrelevant ones may introduce noises and lower the linking accuracy. To address this issue, we design and combine the hyperlink extraction stage and the hyperlink attention stage to learn more suitable hyperlinks in the document-level disambiguation. In addition, we also enhance the context-level disambiguation by utilizing additional entity descriptions and work on retrieving high-quality candidate set for entities at the beginning of our model. Experimental results show that our proposed model outperforms the state-of-the-arts on various benchmark datasets, and even being competitive to the models that rely on additional information.

Abstract: In this study, we propose a new model aiming to enhance the quality of entity linking by exploiting highly relevant hyperlinks in knowledge base for entity disambiguation. We find that most existing studies do not filter the corresponding hyperlinks for each entity, where using the irrelevant ones may introduce noises and lower the linking accuracy. To address this issue, we design and combine the hyperlink extraction stage and the hyperlink attention stage to learn more suitable hyperlinks in the document-level disambiguation. In addition, we also enhance the context-level disambiguation by utilizing additional entity descriptions and work on retrieving high-quality candidate set for entities at the beginning of our model. Experimental results show that our proposed model outperforms the state-of-the-arts on various benchmark datasets, and even being competitive to the models that rely on additional information.

Abstract: This paper presents the VLSI architecture and hardware implementation of a highly efficient deblocking filter (DBF) for High Efficiency Video Coding systems. In order to reduce the number of data accesses and thus to enhance the timing efficiency, novel data structures and memory access schemes for image pixels are proposed. Furthermore, a novel edge-fetching order is presented to strike a balance between the processing throughput and complexity. Based on the proposed structure and access pattern, a six-stage pipelined two-line DBF engine with low-latency data access sequence is designed, aiming to achieve high processing throughput while at the same time maintaining low complexity. The detailed storage structure and data access scheme are illustrated and VLSI architecture for the DBF engine is depicted in this paper.

Abstract: This paper presents an algorithm and a VLSI architecture of a configurable joint detection and decoding (CJDD) scheme for multi-input multioutput (MIMO) wireless communication systems with convolutional codes. A novel tree-enumeration strategy is proposed such that the MIMO detection and decoding of convolutional codes can be conducted in single stage using a tree-searching engine. Moreover, this design can be configured to support different combinations of quadrature amplitude modulation (QAM) schemes as well as encoder code rates, and thus can be more practically deployed to real-world MIMO wireless systems. A formal outline of the proposed algorithm will be given and simulation results for 16-QAM and 64-QAM with rate-1/2 and rate-1/3 codes will be presented.

Abstract: Various eHealth applications based on the Internet of Things (IoT) contain a considerable number of medical images and visual electronic health records, which are transmitted through the Internet everyday. Information forensics thus becomes a critical issue. This paper presents a data hiding algorithm for absolute moment block truncation coding (AMBTC) images, wherein secret data, or the authentication code, can be embedded in images to enhance security. Moreover, in view of the importance of transmission efficiency in IoT, image compression is widely used in Internet-based applications. To cope with this challenge, we present a novel compression method named gradient-based (GB) compression, which is compatible with AMBTC compression. Therefore, after applying the block classification scheme, GB compression and data hiding can be performed jointly for blocks with strong gradient effects, and AMBTC compression and data hiding can be performed jointly for the remaining blocks. From the experimental results, we demonstrate that the proposed method outperforms other state-of-the-art methods.

Abstract: Recently, as one of the most promising biometric traits, the vein has attracted the attention of both academia and industry because of its living body identification and the convenience of the acquisition process. State-of-the-art techniques can provide relatively good performance, yet they are limited to specific light sources. Besides, it still has poor adaptability to multispectral images. Despite the great success achieved by convolutional neural networks (CNNs) in various image understanding tasks, they often require large training samples and high computation that are infeasible for palm-vein identification. To address this limitation, this work proposes a palm-vein identification system based on lightweight CNN and adaptive multi-spectral method with explainable AI. The principal component analysis on symmetric discrete wavelet transform (SMDWT-PCA) technique for vein images augmentation method is adopted to solve the problem of insufficient data and multispectral adaptability. The depth separable convolution (DSC) has been applied to reduce the number of model parameters in this work. To ensure that the experimental result demonstrates accurately and robustly, a multispectral palm image of the public dataset (CASIA), is also used to assess the performance of the proposed method. As result, the palm-vein identification system can provide superior performance to that of the former related approaches for different spectrums.

Abstract: Cerebrovascular atherosclerosis has been identifed as a prominent pathological feature of Alzheimer’s disease (AD); the link between vessel pathology and AD risk may also extend to extracranial arteries. This study aimed to determine the efectiveness of using arterial pulse-wave measurements and multilayer perceptron (MLP) analysis in distinguishing between AD and control subjects. Radial blood pressure waveform (BPW) and fnger photoplethysmography signals were measured noninvasively for 3 min in 87 AD patients and 74 control subjects. The 5-layer MLP algorithm employed evaluated the following 40 harmonic pulse indices: amplitude proportion and its coefcient of variation, and phase angle and its standard deviation. The BPW indices difered signifcantly between the AD patients (6247 pulses) and control subjects (6626 pulses). Signifcant intergroup diferences were found between mild, moderate, and severe AD (defned by Mini-Mental-State-Examination scores). The hold-out test results indicated an accuracy of 82.86%, a specifcity of 92.31%, and a 0.83 AUC of ROC curve when using the MLP-based classifcation between AD and Control. The identifed diferences can be partly attributed to AD-induced changes in vascular elastic properties. The present fndings may be meaningful in facilitating the development of a noninvasive, rapid, inexpensive, and objective method for detecting and monitoring the AD status.

Abstract: Metabolic syndrome (MetS) increases the risk of the subsequent development of cardiovascular disease. Thisstudy aimed to determine if the harmonic indexes of finger photoplethysmography (PPG) waveforms can beused to discriminate different arterial pulse transmission conditions between MetS and healthy subjects. Three-minute PPG signals were obtained in 65 subjects, who were assigned to 3 age-matched groups (MS,with no less than three MetS factors; pre-MS, with one or two MetS factors; Control: with no MetS factor). FDT(foot delay time) and amplitude proportions (Cn) and their standard deviations (SDn) and coefficients of variations (CVn) were calculated for harmonics 1 to 10 of the PPG waveform. FDT was smaller in MS than in Control.C1 and C2 values were significantly smaller, whereas C4–C9 values were significantly or appeared to be larger in MS than in pre-MS. Most of the SDn and CVn values were largest in MS. This study is the first to demonstrate that harmonic-analysis indexes of the beat-to-beat PPG waveform can provide information about MetSinduced changes in the arterial pulse transmission and cardiovascular regulatory activities. The present findings may therefore be useful in developing a noninvasive and easy-to-perform technique that could improve the early detection of cardiovascular diseases.

Abstract: Ultrasound-induced microbubble (USMB) cavitation is widely used to promote drug delivery. Our previous study investigated USMB targeting the round window membrane by applying the ultrasound transducer to the tympanic bulla. In the present study, we further extended the use of this technology to enhance drug delivery to the inner ear by introducing the ultrasound transducer into the external auditory canal (EAC) or applying it to the skull. Using a 3-dimensional–printed diffusion apparatus mimicking the pathway for ultrasound passing through and reaching the middle ear cavity in vitro, the models simulating the transcanal and transcranial approach demonstrated 4.8-fold– and 3.7-fold–higher delivery efficiencies, respectively. In an in vivo model of guinea pigs, by filling tympanic bulla with microbubbles and biotin-FITC, USMB applied transcanally and transcranially induced 2.8-fold and 1.5-fold increases in biotin-FITC delivery efficiencies, respectively. In addition, the gentamicin uptake by cochlear and vestibular hair cells and gentamicin-induced hair cell loss were significantly enhanced following transcanal application of USMB. On the 28th day after transcanal USMB, safety assessment showed no significant changes in the hearing thresholds and the integrity of cochlea. These are the first results to our knowledge to demonstrate the feasibility and support the potential clinical application of applying USMB via EAC to facilitate drug delivery into the inner ear.

Abstract: Minoxidil (Mx) is a conventional drug for treating androgenetic alopecia, preventing hair loss, and promoting hair growth. The solubility of Mx has been improved using chemical enhancement methods to increase its skin permeability over the long term. This study created a new ultrasound (US) contrast agent-albumin-shelled microbubbles (MBs) that absorb chitosan oligosaccharide lactate (COL) and Mx-and combined it with sonication by US energy in the water phase to enhance hair growth while shortening the treatment period. COL and Mx grafted with MBs (mean diameter of 1480 nm) were synthesized into self-assembled complexes of COL-MBs and Mx-COL-MBs that had mean diameters of 4150 and 4500 nm, respectively. The US was applied at 3 W/cm(2) for 1 min, and combined with Mx-COL-MBs containing 0.3% Mx. The diffusion of Mx through the dialysis membrane from Mx-COL-MB during US (US+Mx-COL-MB) was more rapid at pH 4 than at pH 7.4, which is favorable given that the environment of the scalp is mildly acidic (pH=4.5-5.5). In Franz diffusion experiments performed in vitro, the release rates at 18 hours in the US+Mx-COL-MBs and US+MBs+Mx groups resulted in 2.3 and 1.7 times the penetration and deposition, respectively, of Mx relative to the group with Mx alone. During 21 days treatment in animal experiments, the growth rates at days 10 and 14 in the US+Mx-COL-MBs group increased by 22.6% and 64.7%, respectively, and there were clear significant differences (p<0.05) between the US+Mx-COL-MBs group and the other four groups. The use of US+Mx-COL-MB in the water phase can increased the effects of Mx so as to shorten the telogen phase, and also increase both the diameter of keratinized hair shafts and the size of hair follicles without causing skin damage.

Abstract: Abstract Response delay is a critical performance measure for detecting the process shift. Reducing response delay in multivariate process monitoring, especially under big streaming data, is one of many data analysis challenges on smart manufacturing. In the literature, researchers introduced the real-time contrast control chart (RTC) to fast detect the process shift. However, most of the RTCs are underperformed in the big data environment. In this research, a new RTC combined with a stacked long short-term memory network named LSTM-RTC was proposed to monitor big streaming data with short response delay. The LSTM-RTC consists of two LSTM layers, a dropout layer between the LSTM layers, and a dense layer to extend the capability to catch the streaming data's hierarchical representation. A variety of synthesized multivariate normal and bivariate gamma datasets were used to evaluate the proposed LSTM-RTC with other benchmark methods. The experiment results show that the proposed LSTM-RTC outperforms the other benchmark RTC methods with the lowest response delay. The evaluation on two real-world cases also highlights the advantage of the proposed LSTM-RTC on real-world process monitoring applications.

Abstract: This paper proposed a fast metaheuristic method for high-dimensional optimizations problem with only one-control parameter in the setting. Essentially, the innovation of the proposed method is to apply automatic k-means clustering on the initial solutions of symbiotic organisms search to create subpopulations. Only the selected elite solutions in each cluster to interact with one another across clusters in the proposed model. This new elite solution searching process can be considered as a combination of local and global searching based on the solution clusters. The proposed method was compared to six representative methods in 28 benchmark problems and 10 composition problems. Also, the proposed method was also compared with four clustering-based metaheuristic methods. The experimental results show that the proposed model is more efficient in its computation and has a better searching quality. For high-dimensional problems, the performances of the proposed method was compared with the original symbiotic organisms search up to 1000 dimensions. The results show that the proposed method can alleviate the dimensionality effect to produce better solution quality with relatively fast computation.