Sensor systems for supporting and protecting life in the future
- Intelligent Sensing

By exceeding human limits through the use of highly reliable, repeatable sensing functions, real-time sensing networks and artificial-intelligence (AI) data analysis, we are making discoveries and accelerating innovation aimed at better support and protection of human life through advanced methods of diagnostic examination.

In 2015, a United Nations summit established sustainable development goals (SDGs) for high-priority global issues. Of the 17 goals, we are focusing on two particularly relevant goals: "Health and welfare for everyone" and "Urban development for sustainable living." We now striving to achieve these goals through innovative research targeting improved methods for examining the human body.

Let's first look at "Health and welfare for everyone." While modern medicine has made amazing progress, the trends of increasing populations in developing nations and both aging populations and fewer children in industrialized nations pose problems for human health, such as newborn and infant illnesses, infectious diseases, hard-to-treat diseases such as dementia and cancer, and adverse reactions to increasingly complex treatments. To overcome these challenges, leading-edge medicine based on fundamental research will play a crucial role in the fields of biochemistry, practical research, clinical pathology and diagnosis/treatment.

Let's look at pathology and actual efforts being undertaken in this field. Today, there are concerns about a shortage of pathologists needed to handle the increasing number of specimens anticipated in the near future. Innovation is strongly needed to provide pathologists with support for their lab work.

In current diagnostic pathology, workflows are typically based on old-fashioned manual procedures. There is growing momentum, however, for technical innovation through digital methods. Moving forward, the key will be to digitalize diagnostic processes through various approaches and establish workflows suited to digitalization as well as frameworks for related quality management.

Numerous changes will be required for this to happen. First, processes must be changed to stabilize imaging quality. In conventional naked-eye observations of microscopic images, pathologists make mental adjustments for variations in quality from one specimen to the next, so the need for process management regarding image quality has not been emphasized highly. In a digitalized workflow, however, imaging quality must be appropriate for use by sensors, i.e., "machine vision." At the same time, the use of artificial intelligence and information processing is also expected to help reduce workloads for pathologists.

Next, let's look at the field of sustainable urban development. Buildings and transportation systems require thorough safety management supported by high standards of professional ethics. But in recent times infrastructure accidents have been occurring due to the aging of bridges and tunnels, as well as automobile accidents due to overcrowded roads.

Given that the degradation of urban infrastructure over time cannot be avoided, there are high expectations for systems that could help to prevent disasters by monitoring infrastructure on a daily basis to promptly detect any sign of danger.

Identifying signs of hidden dangers requires a cooperative cross-industry framework to enable data to be collected from diverse sources, ranging from real-time inspections (embedded sensors) and regular testing to maintenance and repair. The data then can be evaluated using big data analytics supported with artificial intelligence (AI). By using such methods for the early detection of subtle abnormalities, dangers could be forecast far more accurately than through conventional human-based observations and the information could be quickly communicated to initiate appropriate repairs.

Up to this point, innovative testing concepts have been presented from the perspective of maintaining people's health and preserving the living environment. Another way of describing such technological innovation is Intelligent Sensing.

Conventionally, tasks such as testing, observation and measurement have been handled by people, so data is organized as information that people can recognize. For example, when using an image-based testing device, RGB color images or 30 frames-per-second (fps) video are used on the assumption that they will be viewed by humans. But if this same data were used for AI, the manual task of image annotation (classifying image content) to teach a machine(s) would be exceedingly tedious and it is unlikely that the analytical results would greatly exceed human-derived results.

Since our goal is to accurately understand phenomena through data, in the above examples, for instance, spectral data could be used instead of RGB color images and methods for understanding either high-speed or long-duration phenomena could be used rather than 30fps video. In addition, to analyze the data efficiently, the use of "cleansing" processes for data standardization and homogenization would be helpful. And finally, multi-modal and multi-scale approaches would be needed to understand the phenomena from different perspectives to grasp their true nature fully.

Intelligent sensing networks will become possible through open-minded collaboration among experts who share the common goal of protecting and enhancing society. Let us all contribute to our collective wisdom to realize such a society.