A5100420459- Advanced Memory and Neural Computing
- Total Knee Arthroplasty Outcomes
- Orthopaedic implants and arthroplasty
- Ferroelectric and Negative Capacitance Devices
- Anesthesia and Pain Management
- Cardiac, Anesthesia and Surgical Outcomes
- Advanced Chemical Sensor Technologies
- Laser-Plasma Interactions and Diagnostics
- Neural dynamics and brain function
- Cholesterol and Lipid Metabolism
- Laser-induced spectroscopy and plasma
- Photonic and Optical Devices
- Hip disorders and treatments
- CCD and CMOS Imaging Sensors
- Silicon Nanostructures and Photoluminescence
- Esophageal Cancer Research and Treatment
- Hip and Femur Fractures
- Low-power high-performance VLSI design
- Laser-Matter Interactions and Applications
- Neural Networks and Reservoir Computing
- Analytical Chemistry and Sensors
- Neuroscience and Neural Engineering
- Traditional Chinese Medicine Studies
- Advanced X-ray and CT Imaging
- Electrical and Bioimpedance Tomography
Tsinghua University
2016-2025
Oil Crops Research Institute
2018-2024
Chinese Academy of Agricultural Sciences
2018-2024
The University of Texas Southwestern Medical Center
2024
Fifth Affiliated Hospital of Xinjiang Medical University
2023
Xinjiang Medical University
2023
Hubei University of Science and Technology
2022
Women's Hospital, School of Medicine, Zhejiang University
2019-2022
Institute of Microelectronics
2015-2021
Kunming General Hospital of Chengdu Military Command
2018-2020
With the development of on-chip learning processors for edge-AI applications, energy efficiency NN inference and training is more critical. As dominates consumption [1], [2], [4], [5], reduction paramount importance. Spiking neural networks (SNNs) offer energy-efficient compared with convolutional (CNNs) or deepneural (DNNs), but SNN-based have three challenges that need to be addressed (Fig. 22.6.1). 1) During training, some factors involved in ΔW computation are zeros resulting ΔW=O,...
Lipid synthesis is regulated by the actions of Scap, a polytopic membrane protein that binds cholesterol in membranes endoplasmic reticulum (ER). When ER levels are low, Scap activates SREBPs, transcription factors upregulate genes for cholesterol, fatty acids, and triglycerides. rise, sterol to triggering conformational changes prevent activation SREBPs halting lipids. To achieve molecular understanding how regulates Scap/SREBP machine identify therapeutics dysregulated lipid metabolism,...
Strong K\ensuremath{\alpha} emission is observed in picosecond laser-plasma interactions with high-intensity-contrast, p-polarized, laser pulses. from Si substrates overlaid various thicknesses of Al compared a Monte Carlo simulation. The results show that the hot electrons which deposit their energy solid material have 3-keV temperature and carry 10% incident energy. shifted indicates heated to 35 eV up 5000 \AA{} target surface, consistent preheat electrons.
Reducing learning energy consumption is critical to edge-artificial intelligence (AI) processors with on-chip since dominates consumption, especially for applications that require long-term learning. To achieve this goal, we optimize a neuromorphic algorithm and propose random target window (TW) selection, hierarchical update skip (HUS), asynchronous time step acceleration (ATSA) reduce the power consumption. Our approach results in 28-nm 1.25-mm <inline-formula...
The pedestal (prepulse and postpulse) associated with a chirped-pulse-amplification (CPA) laser is studied. Four components have been identified that contribute to the pedestal. Pulses are spectrally shaped by gain narrowing in frequency-matched, regenerative amplifier, while self-phase modulation avoided. intensity contrast further improved through use of saturable absorber, resulting Gaussian pulses ~0.9-ps duration an exceeding 105:1. Both experimental numerical descriptions these...
Spiking Neural Networks (SNNs) offer a highly promising computing paradigm due to their biological plausibility, exceptional spatiotemporal information processing capability and low power consumption. As temporal encoding scheme for SNNs, Time-To-First-Spike (TTFS) encodes using the timing of single spike, which allows spiking neurons transmit through sparse spike trains results in lower consumption higher computational efficiency compared traditional rate-based counterparts. However,...
AIM: To characterize the alteration and significance of p53 PCNA in cancer adjacent tissues concurrent cancers from esophagus gastric cardia same patient. METHODS: P53 protein accumulation 25 patients with (CC, carcinomas esophageal squamous cell carcinoma adenocarcinoma) were detected by immunohistochemical method (ABC). RESULTS: In CC patients, both (SCC) adenocarcinoma (GCA) showed different positive immunostaining extent (P > 0.05). The rates for 60% (15/25) 92% (23/25), respectively...
To improve the positioning accuracy of implants in Total Hip Replacement (THR) surgeries, a visual-aided wireless monitoring system for THR surgery is proposed this paper. This aims to measure and display contact distribution relative pose between femoral head acetabulum prosthesis during help surgeons obtain accurate position implants. The consists two parts: Sensors Array Measuring System (SAMS) part. SAMS composed sensors array (including an image sensor), signal conditioning circuits,...
Abstract Background This study's initial results revealed significant decreases in treatment‐related esophagitis and pneumonitis cases patients with thoracic esophageal squamous cell carcinoma (ESCC) treated involved‐field irradiation (IFI), compared to elective nodal (ENI). report outlines the long‐term trial results, specifically; overall survival (OS), progression‐free (PFS), metastasis‐free (MFS), locoregional (LRFS). Materials Methods Stage II–III ESCC were assigned randomly, a 1:1...
The sensitivity and selectivity profiles of gas sensors are always changed by sensor drifting, aging, the surroundings (e.g., temperature humidity changes), which lead to a serious decline in recognition accuracy or even invalidation. To address this issue, practical solution is retrain network maintain performance, leveraging its rapid, incremental online learning capacity. In paper, we develop bio-inspired spiking neural (SNN) recognize nine types flammable toxic gases, supports few-shot...
Glucose is the main physiological stimulus for insulin biosynthesis and secretion by pancreatic β-cells. Glucose-6-phosphatase (G-6-Pase) catalyzes dephosphorylation of glucose-6-phosphate to glucose, an opposite process glucose utilization. G-6-Pase activity in islets could therefore be important factor control metabolism and, consequently, glucose-dependent secretion. While has been shown present islets, gene responsible this not conclusively identified. A homolog liver...
This paper presents a four-level pulse amplitude modulation (PAM-4) receiver that incorporates continuous time linear equalizer, variable gain amplifier, phase interpolator-based clock and data recovery, 4-tap direct decision feedback equalizer (DFE) for moderate channel loss applications in wireline communication. A dynamic current-mode logic comparator (DCMLC) is proposed employed the DFE. The DCMLC, which adopts logic, breaks trade-off between bandwidth to Q delay traditional (CMLC)....
The average age of population increases worldwide, so does the number total hip replacement surgeries. Total replacement, however, often involves a risk dislocation and prosthetic impingement. To minimize after surgery, we propose an instrumented prosthesis that estimates relative pose between prostheses intraoperatively ensures placement within safe zone. We create model as ball socket joint, which has four degrees freedom (DOFs), including 3-DOF rotation 1-DOF translation. mount camera...
Here, a simple, efficient, and rapid solid phase extraction-gas chromatography (SPE–GC) method was developed for the simultaneous analysis of free/combined phytosterols in rapeseed their dynamic changes during microwave pretreatment oil processing. First, by comparing different methods extracting from cake, Folch considered to be optimal selected subsequent experiments. Subsequently, extraction validated determining recoveries standards (brassinosterol, campesterol, β-sitosterol cholesteryl...
The enhancement of yield stress NaY/insulating oil electrorheological (ER) fluid is found by adding a liquid crystal additive with higher dielectric constant than that the carrier fluid. increases additive's concentration and saturates at high concentrations. A polarization model has been suggested to understand role additives. theoretical calculations agree our experiments very well. We show small amount can drastically improve even in `dry' ER if conductivity or still much smaller...
As the average age of population goes higher, people undergo hip arthroplasty surgery to replace their stiff and painful damaged joints. To reduce risk factor after we developed a visual aided system in previous works. In order solve integration problems commercial camera with other sensors into femoral head minimize area power consumption, this paper propose CMOS image sensor resolution 200 × specially designed for application which each individual pixel measures around 15 μm size chip...
Atherosclerosis is a chronic disease associated with oxidative stress and inflammatory activation the main underlying trigger for cardiovascular disease. There are many health products in market; order to evaluate effect of these products, this paper, novel lipid profiling platform was established using shot-gun mass spectrum method analysis free fatty acid phospholipids, high performance liquid chromatography coupled spectroscopy mediators triacylglycerol, respectively, serum from male...
The "event-driven" feature of asynchronous circuits enables the to work when and where needed, making it a good alternative design low-power circuits. However, are not widely adopted as consequence lack support by conventional EDA tools. In this article, we propose novel flow implement Click-based bundled-data efficiently down mask layout with To ensure timing correctness, put forward an adaptive delay matching (ADM) method perform accurate static analysis for Compared other toolsets,...
In this paper, we introduce an asynchronous energy-efficient convolutional neural network (CNN) accelerator with reconfigurable architecture including six computing cores, each of which contains 5x5 processing elements. With the dynamically architecture, data path, calculation method, activation function, and pooling way size can be modified according to configurable information for different CNN models. global clock is replaced by local pulse signals from Click An pipeline formed elements...
In this paper, we put forward an energy-efficient configurable asynchronous SNN accelerator for energy-constrained applications, which includes 256 neurons and 131K synapses with 8-bit fixed point weight. To achieve high energy efficiency on-chip learning ability, propose a sparse target propagation (S-TP) algorithm design the Click-based bundled-data circuits. The is implemented in 28nm CMOS process, post place router (post-PAR) simulation results indicate that achieves inference power of...