The cold seems to end today, and spring feels close. When I checked the old Japanese calendar, I noticed we are now in Usui—the season when snow turns to rain and the thaw begins.
In yesterday’s morning newspaper, I read that part of cytology screening may soon be performed by AI.
My wife, who had read the same article, asked me, “So… will AI take over your work in pathology?”
I told her, “Cytology and surgical pathology aren’t the same, so that’s still some way off. But I can imagine us moving in that direction.”
Fields with a strong morphological component—radiology and pathology, for example—often match what AI does well. When I asked an AI about this, I got an interesting suggestion: pathologists might use AI less to confirm whether the diagnosis is correct and more to check whether anything has been missed in the differential diagnosis.
In oncology, genetic testing has already become central in many settings. If rapid, comprehensive genomic analysis becomes routine, it’s possible that the act of “diagnosing the cancer itself” may no longer require a pathologist in the way it does today. Still, I suspect the final evaluation and judgment—the part that carries responsibility—will remain a human task for quite a while.
Evidence-based medicine has aimed to reduce the personal variability of medical practice and to place scientific evidence at the forefront. In that sense, AI-driven diagnostic support can be seen as an extension of that trajectory.
More areas of medicine may be replaced or reshaped by AI than we currently expect. As personal variability fades, diagnostic processes may become increasingly standardized through algorithms with a certain degree of objectivity. From interviews and test data, AI can generate comprehensive diagnostic candidates and suggest treatment options that are close to optimal.
If that happens, the roles within clinical teams may be reorganized—especially in domains where procedures are not the main work. Even in psychiatry, diagnostic support based on conversation data will likely advance.
As for me, I already rely on AI knowledge as an aid in my diagnostic work more often than I would have imagined a few years ago. It reduces the risk of careless omissions, and that reassurance is real.
Not long ago, a senior pathologist I’m close to said, “Pathology has about twenty years left.” I don’t think he meant that pathology will simply disappear. I think he meant that in twenty years, the technology and the workflow we call “pathology” may look completely different from what we practice today.
In some areas, AI has already surpassed human ability by a wide margin. Even so, AI is still a tool. And how we choose to use it—how we integrate it into our judgment and our responsibility—will shape what each medical field becomes.
No comments:
Post a Comment