70/100

Moderate evidence

Cite
More
EmailLinkedInX

ChatGPT-built radiotherapy schedules cut patient gaps by up to 77% in a feasibility study

Study title: Evaluating a large language model for radiotherapy patient scheduling under realistic clinical and operational constraints: A feasibility study

Evidence type
Controlled or validated comparison
Sample
n = 2,043
Sources
1 sourcePeer-reviewed
Published
2 min read

Study title: Evaluating a large language model for radiotherapy patient scheduling under realistic clinical and operational constraints: A feasibility study

A feasibility study in Medical Physics tested whether ChatGPT could build radiation therapy schedules under clinical and operational rules. In a six-week retrospective comparison of 2043 appointments, the AI-generated schedules reduced mean gap by 50.8%-77.1% and raised time consistency from 86.7% to 99.5%-99.9% against the clinical schedule, with no overtime.

Read the study abstract

Background: Radiation Therapy (RT) scheduling requires coordination of clinical constraints and operational objectives. Large language models (LLMs) may provide a natural-language interface for generating candidate schedules. Purpose: To evaluate the feasibility of using an existing LLM to generate RT patient schedules under clinical and operational constraints through simulation and retrospective clinical comparison. Methods: A one-year, three-linear-accelerator (LINAC) simulation included 1400 new patients acr…

The gain

In a feasibility study comparing ChatGPT-generated schedules with a clinical schedule across 2043 retrospective appointments, the AI strategies cut mean gap by 50.8%-77.1% and raised treatment-time consistency from 86.7% to 99.5%-99.9% without overtime.

In brief

The work also included a one-year simulation of three linear accelerators and 1400 new patients across 12 treatment categories. The schedules met all predefined feasibility rules, and adding a continuity objective cut the rate of patients switching machines from 52.4% to 9.4%.

The authors describe the approach as a candidate-schedule generator for human-in-the-loop decision support, not an autonomous scheduler, and say the findings support further investigation.

Main points

  1. The simulation covered one year, three linear accelerators and 1400 new patients across 12 treatment categories modeled after clinical practice.
  2. Across scenarios, 98.6%-98.8% of evaluated fractions stayed within the preferred treatment-time window.
  3. Adding a LINAC-continuity objective reduced the LINAC switch rate from 52.4 ± 2.6% to 9.4 ± 0.6%.
  4. Gap-constrained relaxation reduced daily gap to 47.1 ± 8.5 min and Friday gap to 89.2 ± 18.4 min while keeping overtime low.
  5. In infeasible request tests, the framework identified predefined constraint violations and returned interpretable corrective suggestions.
  6. Heuristic and CP-SAT optimization benchmarks also produced feasible schedules but showed strong objective-weight trade-offs.

The full story

Researchers gave ChatGPT structured prompts containing patient data, the current schedule state, clinical rules and scenario-specific objectives. They tested weekly treatment-time consistency, LINAC continuity, gap-constrained temporal relaxation and handling of infeasible requests in a simulated clinic.

The LLM schedules were benchmarked against heuristic and CP-SAT schedules under two objective settings. All approaches produced feasible schedules, but the optimization methods showed strong trade-offs depending on objective weights.

In the retrospective comparison against a real clinical schedule, every computational approach scheduled the workload without overtime. The LLM strategies shortened gaps and brought time consistency to near 100%, while the authors frame the tool as a human-in-the-loop aid.

Sources

1 on file
  1. Medical Physicsdoi:10.1002/mp.70767
    Peer-reviewed2026-12-01

Talk it over

Three questions to start on, two more if you have time
ChatGPT schedules cut mean gap by 50.8%-77.1% in a six-week retrospective comparison of 2043 appointments. What would you want to see in a live clinic before believing that gain would carry over to real patients?

Start with a question

ChatGPT-built radiotherapy schedules cut patient gaps by up to 77% in a feasibility studyhttps://truvace.com/item/evaluating-a-large-language-model-for-radiotherapy-patient-scheduling-under-real

  1. ChatGPT schedules cut mean gap by 50.8%-77.1% in a six-week retrospective comparison of 2043 appointments. What would you want to see in a live clinic before believing that gain would carry over to real patients?

    Answer in the thread
  2. The authors call the tool a candidate-schedule generator for human-in-the-loop decision support, not an autonomous scheduler. Who should be the person who reviews and approves each schedule, and why?

    Answer in the thread
  3. Adding a continuity objective cut the rate of patients switching machines from 52.4% to 9.4% in the one-year simulation. Which patients or clinic staff might be missing from a simulation of three linear accelerators and 1400 new patients?

    Answer in the thread
If you have more time
  • The heuristic and CP-SAT benchmarks also produced feasible schedules but showed strong objective-weight trade-offs. How would a radiotherapy clinic decide which method to use and who sets the weights?

    Answer in the thread
  • If you were a patient waiting for radiotherapy, which would matter more to you: fewer gaps between sessions, or a consistent treatment time of day? What would you ask the clinic?

    Answer in the thread

What readers say