Forty-seven radiation oncology departments serve the Tohoku region. Fourteen of them have a full-time senior dosimetrist on staff. For the remaining thirty-three, that gap is not an abstraction — it shows up every week as delayed treatment starts, rerouted cases, and physics teams stretched past reasonable capacity.
The Planner Gap in Numbers
We've been tracking this closely, and the math is not comfortable. A typical community radiation oncology department sees 8 to 12 new treatment-plan requests per week. A single dosimetrist working with standard TPS tools can realistically complete 5 to 7 high-quality plans in that same window — factoring in contouring review, optimization cycles, peer QA, and physician sign-off.
That's a structural deficit, not a staffing quirk. And it compounds: when volume spikes after a screening campaign or seasonal referral surge, the queue doesn't clear — it lengthens. Patients in need of prostate, head-and-neck, or breast IMRT wait.
Not by days. By weeks.
Why Outsourcing to Academic Centers Creates Its Own Problems
The conventional answer has been to outsource planning to affiliated academic medical centers — Tohoku University Hospital, Miyagi Cancer Center, and similar institutions. On paper, this works. In practice, it introduces a 5 to 10 day delay between simulation and first-plan delivery, depending on queue depth at the receiving center.
For curative-intent radiotherapy, that window matters. Treatment-start delays affect local control rates in time-sensitive histologies. They extend the anxious waiting period for patients who've already been through surgery or chemotherapy. And they add coordination overhead — faxed CT datasets, phone consultations, re-simulation if the plan doesn't translate to the referring site's linac geometry.
Academic center dosimetrists are also not trying to serve community hospitals as a primary mission. They have their own complex cases, clinical trials, and QA obligations. Community referrals often occupy a lower-priority slot in the queue. That's not a criticism — it's just the reality of how tertiary centers operate.
What AI-Assisted Planning Actually Changes
Here's the thing about AI in treatment planning: the value is not in replacing dosimetrists. It's in changing who can do a first-pass plan to a publishable standard.
In our experience with the AIVOT system deployed in Tohoku community sites, a junior dosimetrist working with AI-generated plan templates can match the quality of a senior-dosimetrist plan on approximately 85% of standard cases. Prostate IMRT. Post-mastectomy breast. Oropharyngeal head-and-neck. These are not simple plans — but they follow patterns that AI models can learn and apply consistently, given adequate training data from peer-reviewed clinical protocols.
The physicist still reviews. The radiation oncologist still approves. The dosimetrist still exercises clinical judgment. The AI compresses the first-optimization phase, where a significant portion of planning time is otherwise spent.
Pilot Data from the Tohoku Region
One regional hospital in Miyagi Prefecture ran a structured pilot comparing pre-AIVOT and post-AIVOT planning workflows over 90 consecutive treatment plans across prostate, breast, and H&N cases. The results were direct:
- Plan approval rate: 93% of AIVOT-assisted first plans were approved without major revision (versus 71% with conventional junior-dosimetrist first pass)
- Time-to-first-plan: Median dropped from 2.5 hours to 40 minutes
- Physician re-simulation requests: Fell from 11% to 4% of cases
Ninety plans is a limited dataset. We're not claiming this generalizes to every department or every case mix. But the direction is consistent with what we've seen across the broader Tohoku pilot cohort, and the 40-minute median is reproducible.
Fact: a 40-minute first-plan cycle allows a two-dosimetrist department to clear a 10-plan weekly queue without overtime. That same workload, without AI assistance, required either extended hours or case transfers.
What AI Does Not Replace
This matters, and we want to be precise about it.
AIVOT-assisted planning is validated for standard curative-intent cases: prostate (including pelvic nodal), breast (including regional nodal), head-and-neck (oropharynx, larynx, early-stage oral cavity). These categories represent roughly 60 to 65% of the new-patient volume at most community sites.
It is not the right tool for re-irradiation cases, where dose accumulation history requires senior clinical judgment that no current AI can adequately model. It is not appropriate for pediatric radiotherapy, where tolerance constraints differ substantially from adult protocols and errors carry disproportionate long-term consequences. And it does not replace senior dosimetrist expertise for stereotactic body radiotherapy (SBRT) cases, where planning margins are narrow and delivery precision demands a level of plan optimization that goes beyond template-based approaches.
We've turned away requests to use AIVOT for spinal SBRT. We'll continue to do that. The clinical evidence base for AI-assisted planning in those contexts isn't there yet, and we're not going to claim otherwise.
Capacity Multiplier, Not Headcount Substitute
Community hospitals in Tohoku are not going to solve the dosimetrist shortage by hiring. The pipeline from graduate programs is not matching the expansion of community radiotherapy services, and senior dosimetrists from academic centers are not relocating to regional hospitals at scale. That's a 10-year problem, if it's solvable at all by workforce development alone.
AI-assisted planning gives the dosimetrists already present at community sites a higher ceiling. A junior dosimetrist who can consistently deliver 93% first-pass approval rates on standard cases is functionally contributing at a senior level for that case category. That changes what a two-person physics team can sustain.
In our data, sites using AIVOT have reduced outsourcing to academic centers by approximately 60% for eligible case types. The cases that do transfer out are the ones that should — complex geometry, re-irradiation, SBRT. The standard cases stay local, which means patients begin treatment faster and at the institution where their oncologist practices.
That's the real outcome. Not a technology story. A care-access story, told in days shaved from treatment-start intervals.
