自働化 · JIDOKA
Automation with a human touch, for healthcare.
Robots for the routine steps of diagnosis, so the answer comes in a day, not a season — and people get more of a clinician’s time, not less.
A machine that stops when something is wrong, and calls a person.
In 1924 Sakichi Toyoda finished the Type G automatic loom. If a single thread broke, it stopped itself.1,2 It didn’t go on weaving bad cloth, and it didn’t need someone watching every thread: it stopped, and a person came to fix the problem. Toyota built its production system on that idea and called it jidoka, “automation with a human touch”.1
The everyday Japanese word for automation is 自動化, roughly “becoming something that moves by itself”. Toyota writes it 自働化 instead: the middle character 動 (“move”) gains the person radical 亻 and becomes 働 (“work”). Both words are read jidōka.3,4 The human is built into the word.
Automation. Something that moves by itself.
Jidoka. Something that works by itself, with a person in it.
That’s how we want medicine automated.
Finding cancer is getting easier. Getting an answer isn’t.
57.3%
of people in BC’s breast screening program (ages 50–69) who needed a biopsy after an abnormal mammogram had a diagnosis within 7 weeks in 2023. The national target is 90%.5
57% down to 16%
At one regional hospital in BC, the share of patients who got a same- or next-day biopsy fell from 57% in 2013 to 16% in 2023. The authors point to a shortage of radiologists: without one, an image-guided biopsy can’t happen.6
36 up to 71 days
At the same hospital, the average time from imaging referral to diagnosis nearly doubled over the same ten years.6
< 1 hour
How long the biopsy itself usually takes.7
Most of that time is waiting, not work. A biopsy takes about an hour; the wait around it takes weeks.
And finding “something” keeps getting cheaper and more common. That’s good news, but every preventive scan that finds something sends another person into the same short-staffed biopsy and pathology queue.
We model the whole pathway, from scan to diagnosis to treatment, and work on the one step that holds everyone up. Today that’s tissue diagnosis: biopsy and pathology. It’s lean manufacturing and the theory of constraints, said plainly: speeding up any other step only makes the line in front of the slow one longer.
Robots take the narrow, repetitive steps: guiding a needle, drawing blood, moving specimens. A trained person supervises, handles the hard cases and has the conversation. Like Toyoda’s loom, when something looks wrong, the machine stops and calls them.
Scan, read and sample in a single visit, with one person who owns your result until you have an answer.
We judge ourselves by deaths averted. Survival rates can rise just because a cancer was found earlier, without anyone living longer.8
From “something was seen” to a tissue answer, in one visit.
One named person follows it through to an answer.
Machines take the routine steps; the time they free goes back to the patient.
Deaths averted per year is the score.
We’re early: researching, modelling the pathway and prototyping. We’re starting with diagnosis in British Columbia. Nothing here is a cleared medical device or a clinical service yet.