Innovation Activities for Testing
In modern society, inspection is performed in a variety of fields, such as life sciences, health care, industry, and infrastructure maintenance. Highly knowledgeable and skilled specialists use inspection devices to carry out their work in the field. We recognize that there are issues common to each field when it comes to this sort of inspection. Namely, the declining number of specialists and the increased difficulty of inspections. These two issues combine to increase the burden on the specialists who are actually responsible for inspection.
Needs in the field are undergoing a substantial transformation as a consequence of these kinds of environmental changes. I.e., rather than a desire for better inspection devices, there is a heightened need for streamlining of inspection processes in their entirety and increasing overall added value while reducing the burden on specialists. Day after day, we have thought more and more about how the new concept of “Intelligent Sensing” can be used to respond to such needs.
The digitalization of image sensing would serve as a jumping-off point. Said digitalization made it possible to teach machines decisions by specialists, which allowed AI (artificial intelligence) to partially substitute for specialists when such decisions are made. This will reduce burdens in the field, including by making the education required to train specialists more efficient.
Also, we are considering teaching machines the rules specialists use to make determinations and extracting the intermediate data that constitutes the rules as evidence data. We think this will enable optimization of objective quantitative data-based inspection itself by comparing data expressing the results of processes with evidence data extracted from rules used to make determinations.
Additionally, this will allow for utilization of multi-modal data integrating multiple inspection methods that are grounded in different principles. We think that such an approach can facilitate streamlining of overall processes and increase added value.
We would like to introduce two examples from the field of clinical medicines that use microscopes of how our initiatives are actually progressing: “digital pathology” and “infertility treatments.”
Pathology is a field of diagnosis in which pathologists makes definite diagnoses. Clinicians collect biotissues suspected to be diseases, and these tissue specimens are sliced, stained and placed onto microscope slides. Pathologists use a microscope to observe the microstructure of these tissue specimens, and make definite diagnoses based on their observations. Clinicians receive these diagnoses and decide on final treatment methods. Pathologists serve an important role in facilitating correct treatment, and from a diagnostic perspective can be described as specialists with a wide range of knowledge and skills concerning diseases.
As is the case in other fields, the reduced number of pathologists, a high-skill specialization, has become a serious issue worldwide in this domain as well. Also, heightened demand for “individual precision treatments” using gene level information and technological advances have placed a greater burden on pathologists by requiring them to deal with new inspection reagents and methods. In response to this state of affairs, joint research conducted with National Hospital Organization Kure Medical Center and Chugoku Cancer Center at Olympus to develop AI-assisted pathological diagnosis software is in progress.
The approach of this research is to teach machines the decisions by specialists that cancer is evident in certain parts of images. A major issue has come to light as this research progresses. Namely, the destabilization of the results of determinations made by AI due to the occurrence of variance and deviation in specimens during the process of dyeing and digital scanning when producing pathological specimens. This has sparked discussions concerning standardization of digital pathology images at specialized agencies.
Additionally, discussions that take the long view and attempt to predict the future state of digital pathology that uses ICT (information and communications technology) and AI have also begun to be held at academic conferences. Olympus is actively engaging with such opportunities to facilitate widespread adoption of digital pathology on-site, and lead the way toward resolution of social problems faced by this field.
Next is about joint research concerning infertility treatments. With a greater number of households without children due to late marriages also a factor, there has been a spike in the number of people in Japan receiving infertility treatments. However, this is also a global trend. Micro-fertilization is a method of assisted reproductive technology, in which sperm and ova are removed from the body, and ova are inseminated by injecting them with sperm under a microscope. Olympus has started joint research with the Jikei University School of Medicine in an effort to use AI to support this micro-fertilization.
The approach of this research has been to teach machines rules for making determinations. It is already known to some extent what sort of sperm has a high rate of fertilization.
Thus we are quantitatively measuring and comprehensively assessing sperm morphology, motility, and the like, and teaching AI criteria for determining which sperm would be predicted to have a high fertilization rate. We are attempting to cause this logic for sorting sperm to evolve quantitatively by checking the results produced by AI against actual fertilization rates.
This time, we introduced innovations in the field of clinical medicine. Similar social problems have become apparent in many fields where professional inspections are conducted.
In order to solve these problems, open minded collaboration with people who share their goals is indispensable. Together with all of you, we will strive to realize this goal.