English
JapaneseAmplifying the number of rare specialists – AI image diagnosis solves the bottleneck in “image interpretation”
Japan has the highest number of CT and MRI machines per capita in the world. However, there is a relative shortage of radiologists capable of interpreting the vast number of images produced.
How can AI untangle this structural imbalance of “abundant equipment, scarce specialists”?
Japan has the most abundant opportunities for medical examinations in the world. However, the time available to specialists who interpret those images is limited – Japanese diagnostic imaging is a place where this abundance and scarcity coexist.
Japan is at the forefront of the world in the widespread adoption of advanced medical equipment such as CT and MRI scanners. With 51.7 MRI machines per million people, it has approximately twice the G7 average (25.8 machines), and also boasts the world’s highest number of CT scanners (OECD statistics). Good access to these examinations for patients is inherently a significant strength. The problem lies beyond that.
Abundant examinations, but difficulty in interpreting images.
No. 1 in the
world
Number of CT and MRI scanners per capita. The number of MRI scanners is approximately twice the G7 average.
8,137
items
The annual potential workload for CT and MRI scans per radiologist is the second highest in the world, after the United States.
Approximately
40%
In Japan, the percentage of CT and MRI scan results interpreted by radiologists. In Europe, many countries have specialists prepare all reports.
Figure 1: The structure is such that “the diagnostic equipment is world-class, but there is a relative shortage of specialists to interpret the images.” The workload per person is among the highest in the world, and specialists are involved in the interpretation of only about 40% of the examinations.
The bottleneck is not “taking pictures,” but “reading.”
Task 1
Imbalance between demand for image interpretation and the number of specialists
The total number of specialists capable of interpreting images is insufficient to match the volume of images that can be captured. As a result, the involvement of specialists is limited to only a portion of examinations.
Task 2
Regional uneven distribution
Studies have shown that the distribution of radiologists varies by region, with a higher number of specialists in certain areas being associated with greater involvement in image interpretation. Securing a robust image interpretation system becomes more difficult in rural areas.
Task 3
Time required to determine the urgency of the situation
Finding a truly urgent case among a large volume of tests is time-consuming and burdensome. Simply processing them sequentially risks delaying the detection of urgent cases.
AI doesn’t “read for you,” but rather “organizes the text beforehand.”
The important point here is not to entrust the final diagnosis to AI, but to allow the time of rare specialists to focus on making the most valuable decisions.
Triage (prioritization)
AI detects findings that are in high urgency and places them at the top of the waiting list for interpretation. This allows specialists to attend to urgent patients first, preventing oversights and delays in emergency cases.
Image interpretation support and prevention of oversights
By automating the identification of potential abnormalities and quantitative measurements, the quality of image interpretation is improved. It acts as a “second eye” for specialists, reducing the burden of verification.
Figure 2: The concept of AI triage. The AI analyzes the tests that pile up in the order of arrival and rearranges them to prioritize urgent findings. This ensures that the limited time of specialists is used for the patients with the greatest urgency. The final diagnosis is always made by a physician.
The value of AI in medical imaging lies not in eliminating the need for specialists, but in enabling a single specialist to treat more patients more safely.
Points to note: Image diagnostic AI as a medical device requires approval and certification under the Pharmaceuticals and Medical Devices Act, as well as appropriate verification and operational systems at each facility. Implementation should be considered not only in terms of “accuracy,” but also in conjunction with existing image interpretation workflows and the design of responsibility boundaries.
Expertise can not only be increased, but also “amplified.”
Given the difficulty of rapidly increasing the number of personnel, the most realistic solution is to amplify the capabilities of the experts currently on the front lines. This approach applies not only to diagnostic imaging but also to the increasingly strained caregiving and monitoring fields in our super-aging society. Next time, we will discuss the shortage of caregiving personnel projected for 2040 and the role of AI in supporting prevention and monitoring.
