Catalogue Search | MBRL
Are you sure you want to remove the book from the shelf?
{{itemTitle}}
5,958
result(s) for
"Lin, Chun-Lin"
Sort by:
Development of revised ResNet-50 for diabetic retinopathy detection
Background
Diabetic retinopathy (DR) produces bleeding, exudation, and new blood vessel formation conditions. DR can damage the retinal blood vessels and cause vision loss or even blindness. If DR is detected early, ophthalmologists can use lasers to create tiny burns around the retinal tears to inhibit bleeding and prevent the formation of new blood vessels, in order to prevent deterioration of the disease. The rapid improvement of deep learning has made image recognition an effective technology; it can avoid misjudgments caused by different doctors’ evaluations and help doctors to predict the condition quickly. The aim of this paper is to adopt visualization and preprocessing in the ResNet-50 model to improve module calibration, to enable the model to predict DR accurately.
Results
This study compared the performance of the proposed method with other common CNNs models (Xception, AlexNet, VggNet-s, VggNet-16 and ResNet-50). In examining said models, the results alluded to an over-fitting phenomenon, and the outcome of the work demonstrates that the performance of the revised ResNet-50 (Train accuracy: 0.8395 and Test accuracy: 0.7432) is better than other common CNNs (that is, the revised structure of ResNet-50 could avoid the overfitting problem, decease the loss value, and reduce the fluctuation problem).
Conclusions
This study proposed two approaches to designing the DR grading system: a standard operation procedure (SOP) for preprocessing the fundus image, and a revised structure of ResNet-50, including an adaptive learning rating to adjust the weight of layers, regularization and change the structure of ResNet-50, which was selected for its suitable features. It is worth noting that the purpose of this study was not to design the most accurate DR screening network, but to demonstrate the effect of the SOP of DR and the visualization of the revised ResNet-50 model. The results provided an insight to revise the structure of CNNs using the visualization tool.
Journal Article
Highly Efficient and Stable White Light‐Emitting Diodes Using Perovskite Quantum Dot Paper
Perovskite quantum dots (PQDs) are a competitive candidate for next‐generation display technologies as a result of their superior photoluminescence, narrow emission, high quantum yield, and color tunability. However, due to poor thermal resistance and instability under high energy radiation, most PQD‐based white light‐emitting diodes (LEDs) show only modest luminous efficiency of ≈50 lm W−1 and a short lifetime of <100 h. In this study, by incorporating cellulose nanocrystals, a new type of QD film is fabricated: CH3NH3PbBr3 PQD paper that features 91% optical absorption, intense green light emission (518 nm), and excellent stability attributed to the complexation effect between the nanocellulose and PQDs. The PQD paper is combined with red K2SiF6:Mn4+ phosphor and blue GaN LED chips to fabricate a high‐performance white LED demonstrating ultrahigh luminous efficiency (124 lm W−1), wide color gamut (123% of National Television System Committee), and long operation lifetime (240 h), which paves the way for advanced lighting technology.
Perovskite quantum dot paper, a new type of perovskite quantum dot film, is demonstrated. Using a simple, fast, scalable, and inexpensive paper fabrication process, the resulting perovskite quantum dot paper is uniform, of high quality, and very stable; it is able to bear high energy radiation and greatly improve the efficiency of perovskite quantum dot–based white light‐emitting diodes.
Journal Article
Similar efficacy and safety between lenvatinib versus atezolizumab plus bevacizumab as the first‐line treatment for unresectable hepatocellular carcinoma
Background
Lenvatinib and atezolizumab plus bevacizumab(A + B) have been used for unresectable hepatocellular carcinoma (HCC) as first‐line therapy. Real‐world studies comparison of efficacy and safety in these two regimens are limited, we therefore conduct this study to investigate these issues.
Methods
We retrospectively reviewed patients received lenvatinib (n = 46) and A + B (n = 46) as first‐line systemic therapy for unresectable HCC in a tertiary medical center. Objective response rate (ORR), progression free survival (PFS), and overall survival (OS) were evaluated according to modified Response Evaluation Criteria in Solid Tumors (mRECIST). Inverse probability weighting (IPW) was performed for baseline clinical features balance.
Results
A total of 92 patients with median age of 63.8 year‐old, 78.3% male, 85.9% viral hepatitis infected, 67.4% BCLC stage C were enrolled. The median treatment and follow‐up duration were 4.7 months and 9.4 months, respectively. There was no significant difference in ORR (26.1% vs. 41.3%, p = 0.1226), PFS (5.9 vs. 5.3 months, p = 0.4066), and OS (not reached vs. not reached, p = 0.7128) between the lenvatinib and A + B groups. After IPW, the results of survival and response rate were also compared. Subgroup analysis suggested that using lenvatinib was not inferior to A + B in regards of PFS, including those with elder, Child‐Pugh class B, beyond up‐to‐seven, or portal vein invasion VP4 patients. Among the lenvatinib treated patients, multivariate analysis showed patients elder than 65‐year‐old was an independent predictor associated with shorter PFS (adjust HR: 2.085[0.914–4.753], p = 0.0213). The incidence rates of adverse events were similar between two groups (76 vs. 63%, p = 0.1740). Both of two regimens had similarly few impact on liver function by comparison of baseline, third month, and sixth month albumin‐bilirubin index and Child‐Pugh score.
Conclusions
The efficacy and safety of lenvatinib are similar to A + B as a first‐line systemic therapy for unresectable HCC.
This study provides real‐world experience of lenvatinib and A + B as firstline treatment for unresectable HCC. The data showed that compared survival, response rate, and adverse events between two regimens.
Journal Article
Portable biosensor for monitoring cortisol in low-volume perspired human sweat
A non-faradaic label-free cortisol biosensor was demonstrated using MoS
2
nanosheets integrated into a nanoporous flexible electrode system. Low volume (1–5 μL) sensing was achieved through use of a novel sensor stack design comprised of vertically aligned metal electrodes confining semi-conductive MoS
2
nanosheets. The MoS
2
nanosheets were surface functionalized with cortisol antibodies towards developing an affinity biosensor specific to the physiological relevant range of cortisol (8.16 to 141.7 ng/mL) in perspired human sweat. Sensing was achieved by measuring impedance changes associated with cortisol binding along the MoS
2
nanosheet interface using electrochemical impedance spectroscopy. The sensor demonstrated a dynamic range from 1–500 ng/mL with a limit of detection of 1 ng/mL. A specificity study was conducted using a metabolite expressed in human sweat, Ethyl Glucuronide. Continuous dosing studies were performed during which the sensor was able to discriminate between four cortisol concentration ranges (0.5, 5, 50, 500 ng/mL) for a 3+ hour duration. Translatability of the sensor was shown with a portable form factor device, demonstrating a comparable dynamic range and limit of detection for the sensor. The device demonstrated a R
2
correlation value of 0.998 when comparing measurements to the reported impedance values of the benchtop instrumentation.
Journal Article
Spatiotemporal manipulation of ciliary glutamylation reveals its roles in intraciliary trafficking and Hedgehog signaling
Tubulin post-translational modifications (PTMs) occur spatiotemporally throughout cells and are suggested to be involved in a wide range of cellular activities. However, the complexity and dynamic distribution of tubulin PTMs within cells have hindered the understanding of their physiological roles in specific subcellular compartments. Here, we develop a method to rapidly deplete tubulin glutamylation inside the primary cilia, a microtubule-based sensory organelle protruding on the cell surface, by targeting an engineered deglutamylase to the cilia in minutes. This rapid deglutamylation quickly leads to altered ciliary functions such as kinesin-2-mediated anterograde intraflagellar transport and Hedgehog signaling, along with no apparent crosstalk to other PTMs such as acetylation and detyrosination. Our study offers a feasible approach to spatiotemporally manipulate tubulin PTMs in living cells. Future expansion of the repertoire of actuators that regulate PTMs may facilitate a comprehensive understanding of how diverse tubulin PTMs encode ciliary as well as cellular functions.
Tubulin post-translational modifications (PTMs) occur spatiotemporally throughout cells, therefore assessing the physiological roles in specific subcellular compartments has been challenging. Here the authors develop a method to rapidly deplete tubulin glutamylation inside the primary cilia by targeting an engineered deglutamylase to the axoneme.
Journal Article
A robust deformed convolutional neural network (CNN) for image denoising
Due to strong learning ability, convolutional neural networks (CNNs) have been developed in image denoising. However, convolutional operations may change original distributions of noise in corrupted images, which may increase training difficulty in image denoising. Using relations of surrounding pixels can effectively resolve this problem. Inspired by that, we propose a robust deformed denoising CNN (RDDCNN) in this paper. The proposed RDDCNN contains three blocks: a deformable block (DB), an enhanced block (EB) and a residual block (RB). The DB can extract more representative noise features via a deformable learnable kernel and stacked convolutional architecture, according to relations of surrounding pixels. The EB can facilitate contextual interaction through a dilated convolution and a novel combination of convolutional layers, batch normalisation (BN) and ReLU, which can enhance the learning ability of the proposed RDDCNN. To address long‐term dependency problem, the RB is used to enhance the memory ability of shallow layer on deep layers and construct a clean image. Besides, we implement a blind denoising model. Experimental results demonstrate that our denoising model outperforms popular denoising methods in terms of qualitative and quantitative analysis. Codes can be obtained at https://github.com/hellloxiaotian/RDDCNN.
Journal Article