A machine can generate a candidate plan, but physicians and peer review select the plan delivered.
Executive brief
Parallel human and machine plans can expose real adoption gaps that retrospective accuracy misses.
AI value · End-to-end planning time
VerifiedMedian 47 hours; 60.1% reduction
Single-center prospective deployment; 61% of ML plans were selected prospectively versus 83% in simulation.
Before
Manually specifies objectives and iteratively tunes a prostate radiotherapy plan. → Reviews the human-generated plan against standardized criteria.
After
Generates an ML plan in parallel with the conventional human plan. → Blindly compares candidates and selects the clinically acceptable plan for delivery.
Human boundary
Treating physicians and peer review retain final plan selection and delivery authority.
Why it matters
A machine can generate a candidate plan, but physicians and peer review select the plan delivered.
Before
How the work ran before the change.
Medical dosimetrist / physicist
Manually specifies objectives and iteratively tunes a prostate radiotherapy plan.
ControlStandard planning system, dose constraints, and peer review.
Treating physician
Reviews the human-generated plan against standardized criteria.
ControlPhysician and peer-review approval; median end-to-end baseline 118 hours.
What changed
A machine can generate a candidate plan, but physicians and peer review select the plan delivered.
Decision rightSelection moves from a fixed rule to the model
After
How the same work runs now.
Random-forest planning system and human planner
Generates an ML plan in parallel with the conventional human plan.
ControlSame clinical criteria and standardized peer review; two independent candidate paths retained.
Treating physician
Blindly compares candidates and selects the clinically acceptable plan for delivery.
ControlPhysician owns acceptance; unacceptable ML plans are discarded for human plans.
Exception path
An unacceptable ML plan is rejected in favor of the human-generated plan.
Decision authority
Treating physicians and peer review retain final plan selection and delivery authority.
| # | Actor | Action | Control |
|---|---|---|---|
| 01 | Medical dosimetrist / physicist | Manually specifies objectives and iteratively tunes a prostate radiotherapy plan. | Standard planning system, dose constraints, and peer review. |
| 02 | Treating physician | Reviews the human-generated plan against standardized criteria. | Physician and peer-review approval; median end-to-end baseline 118 hours. |
| # | Actor | Action | Control |
|---|---|---|---|
| 01 | Random-forest planning system and human planner | Generates an ML plan in parallel with the conventional human plan. | Same clinical criteria and standardized peer review; two independent candidate paths retained. |
| 02 | Treating physician | Blindly compares candidates and selects the clinically acceptable plan for delivery. | Physician owns acceptance; unacceptable ML plans are discarded for human plans. |
Work that left the path
Human role before
Dosimetrists and physicists iteratively generate a plan for physician approval.
Human role after
Planners oversee candidate generation; physicians select and retain treatment authority.
AI role
Generates a candidate curative-intent prostate radiotherapy plan.
Median 118 hoursMedian 47 hours; 60.1% reduction
Prospective 50-patient deployment · 50 retrospective simulations plus 50 prospective patients
Single-center prospective deployment; 61% of ML plans were selected prospectively versus 83% in simulation.
Anti-pattern
Do not publish simulated acceptance as the prospective adoption rate.
Questions
Portability conditions
Reputation risk
high