AI-based Technology to Retrieve Similar Disease Cases in CT Inspections
Previously, for early lung cancer and cases in which abnormal shadows are concentrated in a single location, doctors used technology which searched for similar cases based on CT images. However, for patients of diffusive lung disease such as pneumonia, abnormal shadows spread in a three-dimensional form throughout the entire organ. In such cases, it was necessary for doctors to confirm three-dimensional similarity. Accordingly, it took a long time for doctors to make a judgment.
Recently, Fujitsu Laboratories Ltd. developed CT inspection technology for searching cases with similar three-dimensional spreading of abnormal shadows from among databases of CT images taken in the past. First, the technology uses machine learning to identify abnormal shadow candidates from CT images (Figure (a)). Next, the technology divides the lung into core and peripheral ranges (Figure (b)) by successively estimating the border between the center and periphery based on relatively clear portions of the CT images. Finally, the technology plots a histogram along the body axis in the vertical direction (Figure (c)) for the number of abnormal shadow candidates existing in both the core and peripheral ranges and examines the characteristics of three-dimensional spreading in the abnormal shadows, thereby searching for similar cases.
The technology shortens the diagnosis time for cases which previously required a long time for doctors to make a judgment. Accordingly, the technology is expected to increase the efficiency of clinical procedures.
Technology for Speeding up Transaction Processing on the Blockchain
Blockchain technology makes it possible to achieve systems with superior resistance to tampering while also ensuring high transparency and reliability-all without the need for a centralized manager. Accordingly, there are high expectations for the use of blockchain in various industries such as finance.
Recently, Fujitsu Laboratories Ltd. succeeded in increasing efficiency for the previous bottleneck of communication processing between the application and blockchain infrastructure. By doing so, Fujitsu Laboratories developed two technologies (differential update (DUS) functionality and compound request (CR) functionality) for increasing the speed of transaction processing. Upon installing these technologies in Hyperledger Fabric v0.6.1(Note) and performing measurement, we achieved an increase in speed of approximately 2.7 times in transaction performance compared to previous methods. This technology makes it possible to apply blockchain technology to online transaction systems which require high performance.
(Note) Hyperledger Fabric v0.6.1:
Stable version of open source software framework in relation to blockchain (as of July 5, 2017)
Virtual Machine Control Technology to Improve Server Density of Datacenter Racks
Currently, the number of servers mounted per rack at datacenters is decided by keeping the total value for the server rated power below the power supply amount of the rack. However, there are actually many cases in which the server load is about 10% to 50%. This means that the amount of power used per rack in proportion to load is low compared to the rated power.
Recently, Fujitsu Laboratories Ltd. developed virtual server control technology to achieve efficient server allocation. This technology establishes a partition made up of backup servers in the datacenter and then migrates virtual servers to the backup partition based on the physical distribution and power consumption of the virtual servers. This makes it possible to reduce datacenter space by improving the mounting density for racks with operating virtual servers. In one case with a server rack operation rate of 90%, calculation showed a space reduction of 40%.
Virtual machine control technology to improve server density of datacenter racks
High-Speed Forensic Technology for Grasping the Entirety of a Cyber-Attack at a Glance
Assessing the damaging effects of malware, which infect organizations to trigger problems, used to involve bringing experts in for time-consuming analyses of network and device logs.
By compressing, storing, and automatically analyzing massive volumes of network communications data, Fujitsu Laboratories has now developed technology that lets users analyze the status of a targeted cyber-attack in a short period of time and shows the whole picture at a glance. The technology collects communications data flowing through the network and then, using the data to infer the commands executed on the PC, connects command operations with specific user information to identify who executed what type of remote control and collect trace information about command operations. Upon detecting a targeted cyber-attack in its home system, the technology extracts affected PCs one after the other and automatically draws the attack status from a comprehensive perspective.
As a result, users can conduct analyses quickly without needing to be an expert, and can implement prompt, far-reaching countermeasures before the damage of an attack can spread.
The Industry's First Technology for Automatically Assessing Personal Data Privacy Risks
Japan's Amended Act on the Protection of Personal Information, goes into effect in 2017, will make it possible to provide third parties with de-identified personal data-even without the individual's consent. Before providing de-identified data, providers must first ensure it complies with guidelines and evaluate the risk of specific individuals being recognized.
Based on data distribution, Fujitsu Laboratories developed the industry's first technology to automatically search for combinations of attributes that make it easiest to identify individuals-as well as quantifying that ease of identification-in a realistic timeframe. This makes it possible to quickly see which attributes to prioritize for de-identification. Fujitsu Laboratories also developed a technology for calculating potential damage from data leaks and determining compliance with various de-identification guidelines. Users can evaluate risks and implement appropriate de-identification processes quickly and easily.
Energy-Saving Technology for Cooling Data Centers
Currently, reports indicate that data center energy consumption constitutes 1–2% of all electricity use, so better energy conservation is needed, particularly in air-conditioning.
To respond to the dynamic status changes unique to data centers, such as moving information equipment in and out and modifying rack arrangements, Fujitsu Laboratories has developed a highly accurate prediction technology. The innovation sequentially predicts air-conditioning effects, making it possible to reduce the amount of energy that air conditioning consumes. The technology creates a database that incorporates the status of air-conditioning equipment and operates under conditions requiring the system to gather information from two sources: not only choosing measurement data for addressing prediction targets but also automatically selecting at least one variable from the air-conditioning equipment status. By creating a predictive model using the selected variables, the technology successfully improves predictive accuracy.
Development of in-memory de-duplication technology to speed up the response of large-scale storage
In recent years, performance requirements for storage systems are increasing due to utilization by businesses of big data and the advance of virtualization. For this reason, "all-flash arrays," capable of extending performance by bundling multiple devices in a high-speed network using high-speed flash devices such as SSDs and the like, have become popular. In all-flash arrays, in order to effectively use the limited flash device capacity, de-duplication technology that eliminates duplicate data when writing to the flash device is installed. However, because there is a need to search for duplicate data via the network connecting the flash devices every time data is written, there was a problem that the response speed when writing decreases as the storage devices become larger and faster.
Fujitsu Laboratories Ltd. has developed a new method that can speed up the response speed by performing de-duplication after the data is written. In addition, since data writing to the memory is performed twice when processing according to the new method is continued, the communication volume increases and the overall processing performance is degraded. For this reason, a technology has been developed to switch automatically between the new method and the conventional method depending on the operational conditions. This technology can speed up response speeds by up to two times, realizing improved response of virtual desktop services and faster database processing times.
Developed in-memory de-duplication technology
Efforts to evaluate potential to reduce CO2 from the research and development stage
In April 2010, Fujitsu began efforts whereby researchers evaluate to what extent the advanced technologies that they are working on can contribute to the environment (potential for reduction of CO2 emissions at the usage stage of products and services in which the technologies are incorporated). This has been achieved by applying an environmental contribution assessment tool to all advanced technologies that the company researches and develops. Through these efforts, it has become possible to develop advanced technologies that are better for the environment, contributing to reducing the environmental impact of customers and society as a whole through the products and services in which these technologies are incorporated.
Evaluation of the contribution to the environment of technologies from the research and development stage
Outline of environmental contribution evaluation tool
Estimated data such as reduction in power consumption and man-hours before and after incorporation of the developed technology into the product or service, and usage conditions of the product or service (scale and usage time) is input. These input items are then multiplied by the corresponding specific CO2 emissions to quantitatively calculate the CO2 reduction of the technology (see diagram).
Moreover, in performing the calculation, a roadmap that clarifies the relationship with the commercialization plan for the product or service in which the technology is incorporated is used to make it possible to estimate how the product or service with the technology will be used by customers. In this way, the data required to quantitatively evaluate CO2 emission reductions is identified.
Environmental contribution assessment tool Screen