English
JapaneseCreating “usable” AI : How to design trust and data in the medical field
It worked well in the demo, but it doesn’t last in practice. Many medical AI and digitalization projects hit this “implementation wall.” In this final installment, we will deal not with the technology itself, but with the design required to make it reliable and sustainable.
A great system doesn’t become valuable the moment it’s created, but only when it’s used continuously – the success or failure of technology in healthcare depends on the design of its implementation and operation.
The collaborative support, business process automation, medical imaging, and monitoring we’ve seen in this series all share one common thread: the significant gap between “creating something that works” and “ensuring it continues to be used safely in the field.” Even if effectiveness is confirmed in pilot tests, without a system for operation, maintenance, security, and scalability, the mechanism will quietly become obsolete.
Where are the “implementation hurdles”?
Task 1
Personalized operations
If configuration changes and troubleshooting after implementation depend on a single person, the system will grind to a halt the moment that person becomes unable to work. The more successful a project is, the more likely it is to face this operational risk.
Task 2
Lack of reliable design
If AI’s decisions are a black box, medical professionals cannot use it with confidence. Understanding “why” it made a particular decision, and ensuring that humans retain the final say, are essential prerequisites for its use.
Task 3
Data Protection and Compliance
Medical information is considered “sensitive personal information” under the Personal Information Protection Act. Systems that do not meet the requirements for secure management cannot be put into actual operation in the first place.
Figure 1: There is a gap between a “successful prototype” and “ongoing implementation in the field” in terms of operation, maintenance, security, scalability, and reliability. This gap cannot be bridged if operations rely on a single person.
Two layers of trustworthy medical AI
“Usable AI” requires two layers of design. One is a human-centered design. The other is the data and security infrastructure that supports it.
Human-centered (Human-in-the-loop)
The AI makes suggestions, but the final decision is always made by a human. We incorporate a cycle into the design where the AI improves based on feedback from the results. We clearly define explainability and the boundaries of responsibility.
Data and security foundation
The design will be based on safety management in accordance with the three ministries’ two guidelines (Ministry of Health, Labour and Welfare, Ministry of Economy, Trade and Industry, and Ministry of Internal Affairs and Communications), appropriate protection of sensitive personal information, and consistency with regulations such as the Next Generation Medical Infrastructure Act.
Designed with operation in mind.
We create a system and architecture that avoids reliance on specific individuals and allows for continuous monitoring, maintenance, updates, and expansion. The design is based on the premise of “operating and nurturing,” not simply “building and being done.”
Adaptation to on-site workflows
Prioritizing seamless integration into on-site procedures over high precision; and everyday usability over unused high-end features.
Figure 2: The two-tiered structure of a trustworthy medical AI. The upper layer is a human-centered cycle where “AI proposes, humans make decisions, and learn from the results.” The lower layer is a foundation of security and regulatory compliance in accordance with the three ministries’ two guidelines. If either layer is lacking, it will not continue to be used in practice.
The best medical AI isn’t the most accurate model; it’s one that healthcare professionals can use confidently and consistently every day.
Why Cubastion
Since its founding in 2006, Cubastion has completed over 400 projects, all on time. What we value most is not flashy prototypes, but implementations that are safely and continuously used in the field. Data integration, operational-oriented design, and consideration for security and compliance – our design philosophy, which goes beyond “build and forget,” is supported by our organization in India, Japan, and the United States, as well as our Japanese-speaking team in our Yokohama office.
Beyond this series
Throughout these five sessions, we have consistently maintained the same perspective: technology is not meant to replace healthcare workers, but rather to restore time and peace of mind to both patients and healthcare providers. If this perspective aligns with the challenges your hospital or company is facing, we would love to hear from you.
