CONNECT WITH US
Wednesday 13 January 2021
Gowin embeds Winbond 64Mb HyperRAM to GoAI 2.0
Winbond Electronics Corporation, a leading global supplier of semiconductor memory solutions, today announced that FPGA manufacturer Gowin Semiconductor has embedded a Winbond 64Mb HyperRAM fast memory device in its new GoAI 2.0 machine learning platform.GoAI 2.0 is a complete, new hardware and software solution for machine learning applications. Compatible with the TensorFlow machine learning development environment, GoAI 2.0 is aimed at edge computing applications such as smart door locks, smart speakers, voice-activated devices and smart toys.The hardware component of the GoAI 2.0 platform, the GW1NSR4, is a system-in-package (SiP) featuring a FPGA and ARM Cortex M3 microcontroller for the machine learning application supported by Winbond's 64Mb HyperRAM supplied in known good die (KGD) format.The Winbond DRAM based on HyperRAM technology is ideal for Gowin's target applications, in which the electronics circuit needs to be made as small as possible, while providing sufficient storage and data bandwidth to support compute-intensive workloads such as keyword detection or image recognition. Winbond's 64Mb HyperRAM product has just connected 11 signal pins, so its connections to the host FPGA is minimal - the entire GW1NSR4 SiP has a footprint of just 4.2mm x 4.2mm in a BGA package. The 64Mb memory capacity provided by the Winbond device is sufficient to run both an operating system and to concurrently operate as buffer memory for a TinyML model, or as a frame buffer.The performance specifications of the Winbond's 64Mb HyperRAM include maximum data bandwidth of 500MB/s. It also offers ultra-low power consumption in operating and hybrid sleep modes.Jason Zhu, CEO of Gowin, said "The problem which Gowin Semiconductor has solved with the GW1NSR4 is to pack a high-performance and low-power edge computing engine in a tiny package. The Winbond KGD format and HyperRAM memory technology are ideal for this, because the die can be embedded in the same package as the FPGA, eliminating the need for DRAM as an external component"."Gowin is very happy that we had Winbond as our partner in the development of the GW1NSR4 because its expertise in the KGD package format and its help in integrating the die into the SiP helped us realize a very effective and reliable design in a short time."Winbond's HyperRAM products are available for high-volume production in densities of 512Mb, 256Mb, 128Mb, 64Mb and 32Mb. For more information, go to www.winbond.com.Gowin semiconductor embeds 64Mb HyperRAM DRAM from Winbond
Wednesday 6 January 2021
Realizing the smart factory vision through an AIoT approach
Looking back through the year, smart production can be regarded as a global development trend. Covid-19 outbreak has hit the global manufacturing supply chain. Although the labor shortage has led to the serious decline in production across the industry, the epidemic has driven the demand for "smart manufacturing" and is expected to transform the production model with a vision of the "Smart Factory" of the future.A Smart Factory concept is centred around the confluence of digital transformation and intelligent manufacturing trends. It implements the key value of Industry 4.0 with automation, real-time data, connectivity, embedded sensors and machine learning which will transform the traditional production model into a highly customized and intelligent way. The notable component driving these trends is the AI of Things (AIoT).AIoT is the potential synergy lies in this intersection of IoT (or IIoT; Industrial IoT) with AI. At its core, the crucial characteristic defining the AIoT promise (and therefore the Smart Factory) is the massive proliferation of intelligent, automated decision making capabilities across the IIoT network. Decisions begin with the analysis of data, so enabling data analytics with an AIoT approach can be a logical and attractive starting point with relatively low initial investment.Current data analytics capabilities already possess good ROI in their inherent ability to help improve the productivity, quality, yield, or optimise the cost of producing products. An AIoT-enabled data analytics set-up augments these benefits by generating useful data-driven insights that can be used to help the manufacturing system learn from, be optimized, and generate higher performance, or to help users make better decisions. AIoT represents a fundamental paradigm shift for the industry and a massive step towards the vision of achieving the Holy Grail of zero DPPM manufacturing while boosting quality, yield and cost standards across the entire electronics manufacturing value chain.Let's look at one example. Wire Bonding is one of the critical processes in semiconductor assembly that can generate defects, and quality assurance via sampling to spot defects at the factory gate is a typical step. However, sampling has its risks, as defective components can escape detection, with serious liability issues when lives are put at risk. Root cause analysis is also very onerous when defects are detected, massive data points are typically involved making traditional analysis very difficult. The cost of product recalls can also be very costly.The relentless chase towards zero DPPM is thus understandable but a hard vision to achieve using traditional means. An AIoT-enhanced data analytics approach for the Wire Bonding Process flips this scenario on its head, with quality data points analysed in real time. It provides on time, high integrity data points which are crucial for accurate and timely analysis. Out-of-control process anomalies, and defect generation predicted and automatically corrected before output quality is affected. Predictive, automated decision making such as this - 100% quality assurance without Human Intervention - would totally revolutionize and transform current industrial approaches toward Quality Control.Realizing the vision of zero DPPM, ASMPT's newly developed IIoT platform software- SkyEye is ready to drive the "Smart Factory." SkyEye manages and brings insights and intelligence across your macro and micro automation, process equipment, autonomous inspection and metrology. It revolutionized operation management through its remote functionality equipped with AR and VR capability; for off-site assistance, troubleshooting, recovery and process development.Connectivity to ASMPT AIoT Cloud Services provide End to End Data Analytics Life Cycle & Knowledge Management for your AI enabled manufacturing. With an AIoT approach properly put in place, the entire manufacturing value chain can be increasingly equipped with new ways to develop, innovate, and manufacture. The ultimate goal is to produce faster, better and more cost- effective and flexible, without compromising increasingly stringent standards of safety and quality. Please visit ASMPT official website for full text.ASMPT's newly developed IIoT platform software- SkyEye is ready to drive the Smart FactoryAn AIoT-enabled data analytics set-up augments these benefits by generating useful data-driven
Monday 28 December 2020
Sectors to grow and shortages to watch in 2021
In 2021, there's opportunity for growth in many markets, especially automotive, 5G and IoT, and new component releases hitting the market. However, challenges lie ahead.The 2020 government shutdowns due to COVID-19 hindered global supply chains by either halting or limiting production. Despite most manufacturers finally up and running, another issue looms on the horizon - worsening shortages.Growing markets in 2021 and beyondThere are numerous market segments that will experience growth in the upcoming year, including the rollout of 5G, the rebound of the automotive sector, an increased need for datacenters and increasing reliance on AI and IoT by companies.The automotive industry, specifically, will have a big turnaround in 2021. Though automotive demand took a big hit in early 2020 due to the COVID-19 pandemic, it is now recovering and is forecasted to have strong growth in the second half of the coming year.The main driver of this growth is electric vehicles. Other key drivers of growth will be increasing consumer demand for vehicle comfort, the rising consumption of plastics in automotive production and increasing vehicle customization.The automotive interior materials market alone is expected to experience over 5% CAGR from now until 2026, while the aftermarket segment is predicted to see more than 6.5% CAGR in the same time span.The rollout of 5G is also anticipated to have a huge impact on the semiconductor industry as companies prioritize capacity allocation to 5G production. With the 5G network set as the catalyst to multiple technological revolutions, new devices containing higher performing components will be essential requirements.Initially, the 5G rollout may only lead to an increase in demand for chips with a higher baseline memory requirement, but in the coming years the massive number of new applications will go beyond just smartphones. The industrial application segment alone is predicted to experience 221% CAGR between 2021 and 2023, resulting in $497 million in semiconductor revenue.Other markets expected to continue growing in 2021 include the data mining industry and gaming industry. Ongoing demand in these industries will likely lead to increased demand for graphics cards, though this growth may be impeded by yield issues for Nvidia chips. The numerous recent product launches in the gaming industry has already put a strain on GDDR5 and GDDR6 supply, which is likely to worsen as both industries continue to grow.A hindered semiconductor industry in 2021Though there is rising demand from growing industries, various raw material shortages are leading to tight supply, extended lead times and increased pricing for many manufacturers and on a variety of components. This doesn't appear to have an end date in sight and is only thought to get worse in 2021.The ABF substrate shortage is an example of how one shortage can have a widespread impact on components. As a result of demand growth outpacing capacity expansion by 20% in 2020, rising prices have left customers scrambling to fulfill their production needs, especially for CPUs, GPUs, high-end ICs and peripheral products.These shortfalls are likely to widen as demand for 5G infrastructure, networking and high-end GPU and CPU applications outgrows capacity further. GPUs, for example, are expected to experience a 15-20% price increase in 2021 due to this worsening shortage.Another supply constraint expected to further disrupt the semiconductor market next year is the 8-inch wafer shortage, which originated in 2019. With production capacity currently operating at 99% from increased end-product demand, further pressure will be added in 2021 from the growing 5G industry, automotive industry and IC market.Because of the 8-inch wafer shortage, lead times have been worsening and pricing is on the rise. Richtek, for instance, has reportedly had lead times stretch from 4-6 weeks to 26+ weeks on a wide range of ICs, which are commonly used in computers and consumer end-products.Other manufacturers, such as LG, AUO and Innolux are also beginning to feel the effects of raw material shortages, particularly for glass substrates. With demand for LCD panels remaining strong, the glass substrate shortage has caused pricing increases of up to 70% above original price levels.Many manufacturers are competing for a variety of raw materials, including CPU and GPU makers in addition to logic IC and power management IC manufacturers. Because distributor inventories are under allocation, end-customer supply is very tight for a variety of components, which is likely to continue throughout 2021.The takeawayThere are many opportunities for growth in global supply chains going into 2021 created by the rollout of 5G and IoT, the return of the automotive industry and growing demand in data mining and gaming industries. Though rising demand should result in booming industries, numerous shortages could stunt growth and leave manufacturers fighting over resources.
Monday 21 December 2020
Orbotech introduces new flex PCB manufacturing solutions
Orbotech, a KLA company, has announced two new roll-to-roll (R2R) manufacturing solutions for flexible printed circuit (FPCs), enabling the design and mass production of new generations of electronic devices, including 5G smartphones, advanced automotive and medical devices.Orbotech's roll-to-roll solutions for direct imaging (DI) and UV laser drilling overcome many of the yield, throughput and quality challenges inherent in flex material manufacturing. Leveraging newly developed and field-proven technologies, the solutions facilitate high quality, cost-effective mass production of ultra-thin flexible printed circuits that are critical in advanced electronics.The two new series of solutions are the drum-based Orbotech Infinitum for R2R direct imaging, and the Orbotech Apeiron for flex R2R and sheet-by-sheet panel UV laser drilling."Based on the insights gained from nearly 40 years of working with leading manufacturers around the world, Orbotech has created breakthrough technologies and solutions that turn designers' dreams into reality," said Yair Alcobi, president of the PCB division at Orbotech. "Orbotech Infinitum and Orbotech Apeiron build on our existing solutions for flex PCB manufacturing to solve the most pressing challenges faced by advanced flex PCB manufacturers today."New and future advanced electronics with their light weight, smaller form factor and higher functionality increasingly employ delicate flex materials. Both of Orbotech's new solutions optimize the handling of the most delicate flex materials during direct imaging or UV laser drilling. and provide more flexibility to increase production with options for roll widths of 260mm and up to 520mm.Additionally, PCB manufacturers gain significant efficiencies with the smaller footprint of both solutions: smaller clean rooms are required, production capacity per square meter is increased, and power consumption is reduced. These advances create the opportunity for greener flex PCB manufacturing.Orbotech Infinitum
Wednesday 16 December 2020
High performance and low power consumption give Winbond's 1Gb LPDDR3 DRAM the edge in Tsing Micro's new AI image-processing SoC
Winbond Electronics Corporation, a leading global supplier of semiconductor memory solutions, announced that a high-performance, low-power Winbond 1Gb LPDDR3 DRAM product has achieved another great success in high bandwidth memory with Tsing Micro latest artificial intelligence (AI) system-on-chip (SoC).Beijing, China-based Tsing Micro's TX510 SoC is a highly advanced AI edge computing engine optimized for functions such as 3D sensing, face recognition, object recognition and gesture recognition. Paired with Winbond's LPDDR3 DRAM, which offers maximum bandwidth of 1866Mb/s and operates from a dual 1.2V/1.8V supply with power-saving features such as Deep Power-Down mode and a Clock Stop capability, the TX510 offers outstanding speed and accuracy in AI imaging applications.Tsing Micro (www.tsingmicro.com) has implemented an innovative architecture in the TX510 SoC: it includes a 32-bit RISC processor, a reconfigurable neural network engine, a reconfigurable general computing engine, an image signal processor and a 3D sensing engine. For shipping to customers, the TX510 is combined with Winbond's 1Gb LPDDR3 DRAM die in a single 14mm x 14mm TFBGA System-in-Package (SiP).Achieving computing throughput of up to 1.2 TOPS (tera operations per second), this SiP can perform accurate face recognition (false acceptance ratio of 1 in 10 million) in less than 100ms, and compare features with a library of 100,000 faces in less than 50ms. Peak operating power consumption is just 450mW, and the chip uses just 0.01mW in quiescent mode.The TX510 is intended for use in applications which require high-speed image detection and recognition, including biometric sensing, video surveillance, smart retail operations, smart home automation and advanced industrial automation.Wang Bo, CEO of Tsing Micro, said, "Evaluation of the Winbond LPDDR3 DRAM shows that it is highly competitive both in terms of speed and power consumption. But equally important to us when developing the TX510 was the support and expertise that Winbond provided to us in integrating the DRAM die into a SiP in such a way that it maintained high performance and high signal integrity while taking care of thermal management.""Winbond is fast gaining a reputation for providing excellent support to chip and SoC manufacturers which are developing leading-edge AI products. With Winbond low-power DRAM inside, AI products such as the TX510 can set new benchmarks for high performance and throughput while operating from a limited battery power supply," said Winbond.Winbond's LPDDR3 DRAM is in mass production. More information can be found at www.winbond.com.Tsing Micro ships its TX510 AI processor with a Winbond 1Gb LPDDR3 die integrated into a single SiPTsing Micro TX510
Wednesday 9 December 2020
The release of Filecoin could mean product shortages in the semiconductor industry
Years after the release of Bitcoin, the semiconductor industry is in for a splash with the upcoming release of Filecoin, a block-chain based digital storage system. With many users impatiently anticipating Filecoin's release, pent up demand for Filecoin mining services and products will likely translate to a spike in demand for relevant components, such as processors, memory and storage.Today, most data is stored on a centralized cloud storage system hosted by providers, such as Amazon and Google. Because these companies use a central cloud owned system to store everyone's data, it becomes vulnerable to data breaches and outages, such as with the 2017 iCloud hack.Filecoin, on the other hand, builds on InterPlanetary File System (IPFS) and relies on miners to provide storage on a decentralized network. By utilizing a peer-to-peer storage system instead of cloud storage platform, data is protected from a central point of failure, rising storage costs and monitoring.In addition to providing increased storage protection, Filecoin was created to be more accessible to the public, meaning even users without mining experience can contribute to the system easily. In return, contributors will be rewarded with Filecoin cryptocurrency (FIL), incentivizing others to store IPFS content as well.Crypto mining activity, in general, has only continued to increase over the years. Bitcoin, for instance, gained 15 million users in 2019 alone as BTC price remains above US$10,000. Because total FIL is limited to about two billion units, its value will also likely continue to increase - especially as more users join.Unlike with Bitcoin mining, Filecoin mines with hard drives instead of specialized machines. Thus, owning more storage does not result in more effective mining power. To gain more, miners must commit a higher amount of sealed data to Filecoin's system, resulting in a need for powerful processors and high-density HDDs."With the opportunity for anyone to earn FIL, manufacturers that produce DIMMs, CPUs, HDDs, GPUs and motherboards, should expect a spike in demand as users look to gain more mining power. HDD and RDIMM makers, especially, are expected greatly benefit from the upcoming release of Filecoin," states Kingsley Yao, Regional Account Manager, Asia at Fusion Worldwide.Manufacturers may experience more orders for selected products, including newer versions of CPUs, GPUs, RDIMMs and eventually any components used in new mining machine designs. And, since high density HDD is necessary for an IPFS, increased demand could lead to shortages on and higher prices for HDDs and certain CPUs once mass production starts.Those wanting to store their data on Filecoin will also greatly benefit from its release as well. Apart from the greater security and privacy gained from Filecoin's encryption and 'sharding' process, the system's storage cost is also transparent and negotiable. The cloud storage industry is bound to also be affected as competitive pricing will be needed to compete with this decentralized peer-to-peer system.Though Filecoin has only released its "Testnet Incentives" program, numerous Chinese companies have already started selling cloud mining contracts and physical hardware. Multiple mining hardware manufacturers have already reported achieving millions of dollars in sales each tied to Filecoin. Currently, effective mining hardware is predicted to cost more than US$30,000 per set.In the long run, the benefits of mining and storing data on Filecoin will lead to a spike in demand for storage and key components used to build and sell FIL mining machines. Not only will this heightened demand increase pricing for these products, but the cloud storage market may never be the same. As traditional storage data provider giants compete with companies like Filecoin, other industries may need to prepare for the rise of other cryptocurrency powered programs in their markets as well.Filecoin processesPhoto: Company
Friday 4 December 2020
ProteanTecs to host live webinar on SoC quality and time-to-revenue
ProteanTecs, a global leader of Deep Data solutions for electronics' health and performance monitoring, announced today an upcoming live webinar titled "Increasing SoC Quality and Time-to-Revenue with Universal Chip Telemetry (UCT)" which will be hosted on December 9, 2020 at 12pm EST.This webinar will discuss Universal Chip Telemetry (UCT) for chip production, providing high coverage deep data monitoring for parametric visibility in product bring up and volume testing.Attendees will learn:*How to reach volume production faster with higher confidence*How to achieve a 10X DPPM reduction at a fraction of the cost*How to shift-left production for significant cost savingsThe event will feature a live Q&A session."This webinar is a must for anyone dealing with challenges associated with SoC design, bring up, characterization and volume manufacturing," stated Raanan Gewirtzman, CBO. "UCT has the power of providing visibility in all phases of production with apples-to-apples correlation between design and silicon, and across all test stages. When dealing with advanced technologies, companies can't afford to not take advantage of the newly available deep data."To join this free event, register here.About ProteanTecsProteanTecs develops revolutionary Universal Chip Telemetry (UCT) for electronic systems throughout their entire lifecycle, increasing their performance and reliability. By applying machine learning to novel data created by on-chip UCT Agents, ProteanTecs provides meaningful insights and visibility unattainable until today, leading to new levels of quality, reliability and scale. Founded in 2017, the company is headquartered in Israel with offices in New Jersey, California and Taiwan.ProteanTecs live webinar
Thursday 3 December 2020
Exascend introduces industrial rugged PCIe Gen3 and SATA3 SSD products
Exascend Co. Ltd., a leading service-driven high-performance storage solution provider, has announced the availability of new line of thermally optimized, ultra-capacity, high speed and PCIe Gen3 NVMe and SATA3 SSD products based on Marvell's industry leading SSD controller.Exascend PI3 (PCIe Gen3x4) product series are Industrial grade (-40 degree C - 85 degree C operating temperature) SSD with enterprise performance features (sustain high speed and low latency) available in standard U.2 up to 8TB, M.2 2280 up to 4TB. Legacy part such as SI3 Series (industrial SATA3) are still available with capacity up to 4TB and will be a long-term support item.Design and optimized for industrial and rugged applications, where unpredictable operating environment, wide temperature fluctuations, thermal and airflow restrictions pose significant design challenges for small form factor, especially when SATA3 is migrating to high performance PCIe NVMe SSDs."The new generation of industrial applications requires lower latency, consistent QoS, and ever increasing sustained high performance under all operation conditions. PI3 SSD product innovates by fusing Exascend enterprise performance storage with proprietary thermal management technology and extended temperature support. With the introduction of our rugged PCIe SSD for rugged applications," said Exascend CEO Frank Chen, "Exascend is very proud in achieving a significant milestone for the global storage industry. The accelerating trend of Flash adoption for rugged applications, industrial automation, edge computing has called upon service focused companies like Exascend to deliver enterprise performance solutions under wide temperature operating conditions - two conflicting requirements by conventional perceptions. By leveraging our deep expertise in SSD technology, Exascend is able to accomplish both requirements simultaneously and advance our mission to push the boundary of possibilities."Hardware-based AES 256 encryption engine with TCG OPAL 2.0 supportBoth PI3 and SI3 series products support TCG Opal 2.0. By encrypting the entire drive, users do not need to worry about their data being accessed if the drive, laptop, or mobile device is stolen or lost.In addition to standard PCIe NVMe and SATA3 features set, Exascend products are engineered with innovative features and optimizations based on proprietary technologies for rugged applications.Exascend adaptive thermal controlExascend's patented algorithm continuously monitors drive thermals, and regulates drive performance to improve SSD reliability, simultaneously enhancing sustained performance and quality of service (QoS) by 10X. Below is an example of performance benchmark in thermal chamber simulating operation temperature under wide temperature range.4TB M.2 write performance (MB/s)M.2 is far more challenging as opposed to 2.5" or U.2 form factor due to thermal-dissipation is very limited in such small form factor, the higher the capacity, the more difficult it gets. Nevertheless, Exascend's 4TB M.2 can provide a 1,450MB/s sustained write under -40 degree C – 75 degree C, 700MB/s sustained write at 75 degree C – 85 degree C.Exascend accelerated bootWe recognize system responsiveness is a critical aspect of user experience, high capacity Flash storage may take minutes to respond after power cycle, negatively impacting responsiveness. Accelerated boot is our unique technology, by streamlining initialization to prioritize system response, and layering initialization to allow background error correction and recovery, reducing boot time from minutes to seconds, dramatically improving system responsiveness.Exascend rugged feature suiteExascend's standard rugged feature suite including Secure Firmware Update, Tamper Proof Firmware Encryption, Firmware Rollback Protection, Optional Power Loss Protection, pSLC Support, Industrial Temperature Range, Enterprise Performance, High QoS, Low Latency, Fix Major BOM ServiceExascend deep customization serviceExascend's platform-based technology enables deep customization to drive value-add services. We provide engineering support on further performance, power, thermal optimizations and security, vendor unique features on PI3 / SI3 products.Optional value-add featuresExascend's Optional Value-Add Features including thermal, performance and power budget tuning, configurable drive over capacity and provisioning, configurable DWPD (drive writes per day) for extended endurance and life, customizable LED status lighting, customized drive SMART status and telemetry data, write protect or read-only mode for security purpose, integration support of drive life monitoring, rugged erase trigger with quick erase and sanitize erase, custom form factor and connector interface, and exclusive factory data recovery service.Target applicationsPI3/SI3 SSD products are implemented in: factory automation, IoT gateway, transportation, edge server, medical equipment, telemetry devices, 5G telecommunication, autonomous driving, and surveillance.Exascend is a service-oriented provider of innovative standard and custom storage solutions, specialized in low power, high performance, and high reliability products. Since its founding, the company has been awarded more than 45 US and worldwide patents on storage related technologies. "Engineering Inspiration to Innovation" Exascend takes pride in enabling its global customers to push the boundary of possibilities and to differentiate with quality, reliability, and flexibility. For more information about Exascend, please visit www.exascend.com.
Friday 27 November 2020
NCKU team develops digital imaging solutions in liver pathology leveraging AI technology
The pathologist, specialized in looking at tissue or pathological sections through a microscope, is to make a definitive cancer diagnosis and may also work with physicians or other members of healthcare team to recommend a treatment strategy that could include observation, surgery, chemotherapy, radiation therapy, or a combination of these approaches. The pathologist is indispensable to the cancer care team.The interpretation of various pathological sections need to use different tissue marking dyes to find more details of each disease. The regular process will need to use different lenses to magnify objects for carefully observation in viewing macro or micro structure under microscope for a pathologist. It will consume lots of time to identify, count and examine imaging data including Liver fibrosis, Tumor, Hepatic necrosis area, liver portal vein infiltration, and fatty liver detection.Pau-Choo Chung, professor of Department of Electrical Engineering, National Cheng Kung University (NCKU), leads the team to develop digital pathology using image-based solution for the acquisition, management and interpretation of pathology information supported by computational techniques for data extraction and analysis through artificial intelligence (AI) technology. Her pioneer projects are aiming to build an assistant digital pathology system for helping pathologists to save time and efforts to do jobs.The single digital image file of pathological sections is huge sized data sets ranging from 10 to 50 gigabytes. It is necessary to transform the big image file to several smaller and regional image files for further processing. To perform AI training will require massive compute power. Those big data sets require huge GPU computing power and memory allocation. Through the support by National Center for High-Performance Computing (NCHC), this project gets support to access Taiwan Computing Cloud (TWCC) platform to perform the AI Training. The current stage for doing the AI model development, the data set is accumulated around 4000 images by the collaboration with National Cheng Kung University Hospital (NCKUH) and Kaohsiung Veterans General Hospital.For expanding to more use cases after performing AI training process, the project team uses TWCC cloud services to get increases in inference efficiency to fit the requirements of time constraints in hospital. However, it is requiring to link to the service of TWCC for leveraging the AI models under NCHC reinforced cyber-security protection. Some of the hospitals worry patient's data breach, there is another solution to build compute and storage servers inside the hospital's IT infrastructure. Currently the NCKUH and Kaohsiung Veterans General Hospital join force together to deploy this AI system into use as the standard operation procedures. Professor Chung welcomes these investments. And this will help the AI assistant system having further use in the hospitals.The project teams of Professor Chung are heavy users of TWCC services especially in AI training process. The data sets are huge objects to be uploaded to the cloud data center servers. That's the reason why both project team members and TWCC technical support team members frequently to conduct meeting having discussion of ideas for better usage of the TWCC services. It is a great appreciation to see the huge time saving for the project team members.The next steps of project teams are pushing the cloud services of liver pathology assistance systems under carefully considering cyber security priorities. This will help to expand more usage to all Taiwan hospitals with local storage of sensitive medical data. The entire design is aiming to apply TFDA license as a key milestone for making it commercially available in Taiwan. For achieving this goal, it is very important to collaborate with TWCC services. The TWCC service is an invisible hero, said Professor Chung with highly appreciation.Pau-Choo Chung, professor of Electrical Engineering Department, National Cheng Kung University
Friday 27 November 2020
Pulxion's PulStroke device provides an early detection before stroke
Stroke is one of leading causes of death worldwide. Ischemic stroke, caused by the blockage in arteries that supply blood to the brain, is one of the two major types of stroke accounting for 85% in all cases. Carotid artery stenosis is an early warning indicator of ischemic stroke. Carotid ultrasound is the current gold standard to assess the stroke risk. However, it requires medical professionals to operate the equipment, not to mention the precious time spent for appointments and waiting. The entire process is money-costing and time-consuming.For reaching an effective solution, Pulxion Medical Technology, a new start-up company, was spun off from Professor Hao-Ming Hsiao's lab at National Taiwan University. The mission of the company is to develop AI-based imaging technology, with the goals of translating the research work into clinical solutions and commercial products. Thomas Hsu, CEO of Pulxion, highlights that the company uses its core technology of dynamic video processing to build a non-invasive device for stroke risk assessment. The entire process is completed by taking a short video clip aimed at the neck under LED light with only one simple click, anywhere, anytime. In less than five minutes, the user receives a report indicating low to high risk for stroke assessments. With this device, Pulxion plans to promote its use at chain pharmacies and clinics, healthcare centers, and local communities for general public who might have stroke issues without foreseeable symptoms. Early detection is the key to the preventive medicine, Hsu said.Pulxion has collaborated with medical teams from National Taiwan University Hospital (NTUH) for more than five years, with more than 500 clinical cases enrolled so far. These clinical data were used to develop the core technology of dynamic video processing, feature extraction, and AI algorithms. The accuracy of its product stands higher than 90%, when compared to the Carotid ultrasound clinical outcome. The company is planning to apply for regulatory approval from FDA and TFDA to make PulStroke commercially available.In the second development phase of this product, the AI technology will play an essential role for processing massive video imaging data with the trained AI network. In this development phase, Pulxion chooses to enroll in Taiwan Computing Cloud Service (TWCC) from National Center for High-performance Computing (NCHC). The goal is to use parallel computing power of TWCC services to handle massive data for faster and higher efficiency. The use of TWCC powerful computing services will also help to keep all data in Taiwan local servers. This could facilitate the TFDA regulatory approval which is heavily regulated by the cyber-security acts. Applying TWCC services not only increase the high-performance computing power, but also help to meet the TFDA requirements.The design concept of Pulxion's PulStroke device is quick, accurate, and affordable and can be done anywhere, anytime, without the presence of experienced operators. The mission of the company is to provide personalized mobile hospital, allowing general public to assess their own stroke risk at home and take necessary actions as early as possible.Thomas Hsu (back row, second from right), CEO of Pulxion Medical Technology, and the team members