Radiation Oncology Workflow Simulator

Model radiation oncology workflows before changing them in the clinic.

Radonc Workflow Sim lets users build a visual process map, assign staffing and capacity assumptions, configure routing and rework, simulate patient flow, and identify queues, delays, utilization problems, and operational bottlenecks using synthetic data.

Demonstration placeholder
Insert a short GIF here showing a workflow graph on the canvas, a simulation run with animated patient tokens, and the bottleneck / utilization summary. Do not commit a fabricated image.

This local React + TypeScript application (radonc-bottleneck-sim) builds and edits the workflow directly on a tldraw canvas as a directed graph of stages and paths. A deterministic simulation engine routes patients through that graph and replays the run as animated patient tokens.

This is a planning and operations simulation tool only. The defaults in this repo are configurable starting assumptions, not universal clinical truths.

What it does

• The canvas is the source of truth for the workflow: add, remove, connect, and configure stages as custom node shapes
• Each stage node exposes its capacity (units × FTE × hours/day), service time, due date, complexity scaling, and complexity restrictions
• Paths between stages carry a % open routing weight — patients leaving a stage pick among open paths proportionally (a 50% path gets half the flow of a 100% path) — and optional complexity restrictions
• Splitting a stage across multiple nodes (e.g. two planner teams) and weighting paths models partial FTE staffing
• Rework loops are backward paths with low % open
• A Patient Arrivals circle node sets the arrival rate and is where animated tokens enter; a Treatment node books fraction courses against linac capacity
• Simulated runs replay on the canvas: queues stack above each stage, in-service tokens sit inside nodes, overdue work is ring-highlighted, with play/pause, replay, speed, and scrubbing
• Tracks throughput, cycle times, waits, overdue rates, utilization, WIP, machine occupancy, and a bottleneck ranking per stage

Commands

• npm install
• npm run dev
• npm run build
• npm test
• npm run lint

Project structure

src/
  app/         React shell, scenario panel, playback controls, summary
  model/       domain types, default graph scenario
  simulation/  seeded RNG, patient generation, graph routing engine, scheduler, metrics
  analysis/    Little's Law, capacity math, playback frame computation
  nodes/       custom tldraw node shapes (start / stage / treatment)
  ports/       port rendering and pointer interactions
  connection/  connection shape, bindings, insert-on-connection
  components/  toolbar, on-canvas picker, connection edit panel
  tldraw/      canvas seed/extract commands, playback projection, token shape
  tests/       vitest coverage

How a run works

1. On first load the default referral-to-treatment workflow is seeded onto the canvas.
2. Edit the graph: drag from a node's right port to draw a path, click a path to set % open and allowed plan types, edit capacity numbers on nodes, add stages from the toolbar or by dropping them into a connection.
3. Run simulation extracts the graph from the canvas, simulates it deterministically with the configured seed, and replays the result as animated tokens.

Validation status

The repo is intended to validate locally with:

• npm run build
• npm test
• npm run lint

If dependency installation is required, run npm install first.

Clinical-use notice

This software is provided for research, education, and development. It is not a medical device and has not been validated for clinical decision-making.

Users are responsible for independent code review, testing, commissioning, verification of calculations and outputs, and compliance with applicable institutional policies before using any portion of the software in a clinical environment.

No patient information or protected health information is included in this repository. Examples and test data are synthetic or de-identified unless explicitly documented otherwise.

License and credits

MIT — see LICENSE.md.

The interactive node/port/connection canvas machinery is adapted from the tldraw workflow starter template (MIT, © tldraw Inc.). This project is a simulation and planning tool only; it is not a medical device and must not be connected to clinical systems.