run-optimization

Run Optuna parameter optimization for Adaptive Brush segmentation algorithms.

Usage

code

/run-optimization [config.yaml] [--trials N] [--no-exit]

Where:

•config.yaml - Path to YAML config (default: configs/quick_test.yaml)
•--trials N - Override number of trials
•--no-exit - Keep Slicer open after completion for inspection

What This Skill Does

•Loads Slicer with the optimization script
•Loads sample data and gold standard segmentation
•Runs N trials with Optuna TPE sampler and HyperbandPruner
•
For each trial:
- •Applies suggested parameters to the effect
- •Paints all clicks from the gold standard
- •Computes Dice coefficient after each click (for pruning)
- •Reports final Dice as objective value
•Saves results and generates lab notebook

Prerequisites

•Slicer path set in .env file: SLICER_PATH=/path/to/Slicer
•Gold standard exists in SegmentEditorAdaptiveBrushTester/GoldStandards/
•Config YAML exists in SegmentEditorAdaptiveBrushTester/configs/

Execution Steps

Step 1: Read .env and Launch Slicer

bash

source .env && "$SLICER_PATH" --python-script scripts/run_optimization.py configs/quick_test.yaml

Step 2: Monitor Progress

The script outputs progress to terminal:

code

14:30:01 [INFO] optimization: Loaded config: Quick Test
14:30:05 [INFO] optimization: Loaded gold standard: MRBrainTumor1_tumor with 5 clicks
14:30:06 [INFO] optimization: Starting optimization: 10 trials
14:30:08 [INFO] optimization: Trial 0: {'algorithm': 'geodesic_distance', 'edge_sensitivity': 50, ...}
14:30:08 [INFO] optimization:   Click 1/5: Dice=0.3421
14:30:09 [INFO] optimization:   Click 2/5: Dice=0.5234
...
14:30:12 [INFO] optimization:   Final Dice: 0.8734 (4200ms)

Step 3: Review Results

Results are saved to:

code

optimization_results/<timestamp>_<config_name>/
├── config.yaml              # Copy of config used
├── optuna_study.db          # SQLite database (resumable)
├── results.json             # Full results with all trials
├── parameter_importance.json # FAnova importance scores
└── lab_notebook.md          # Human-readable summary

Available Configs

Config	Trials	Purpose
`quick_test.yaml`	10	Fast testing (random sampler)
`default.yaml`	50	General optimization
`tumor_optimization.yaml`	100	Thorough tumor optimization

Output Format

results.json

json

{
  "config_name": "Quick Test",
  "n_trials": 10,
  "best_trial": {
    "trial_number": 7,
    "params": {
      "algorithm": "geodesic_distance",
      "edge_sensitivity": 50,
      "threshold_zone": 50
    },
    "value": 0.8734,
    "duration_ms": 4200
  },
  "parameter_importance": {
    "algorithm": 0.45,
    "edge_sensitivity": 0.30,
    "threshold_zone": 0.15
  }
}

lab_notebook.md

Human-readable markdown with:

•Summary statistics
•Best parameters
•Parameter importance table
•Top 5 trials
•Complete trial history

Examples

Quick test run

bash

source .env && "$SLICER_PATH" --python-script scripts/run_optimization.py \
    SegmentEditorAdaptiveBrushTester/configs/quick_test.yaml

Full optimization with 100 trials

bash

source .env && "$SLICER_PATH" --python-script scripts/run_optimization.py \
    SegmentEditorAdaptiveBrushTester/configs/tumor_optimization.yaml

Override trials and keep Slicer open

bash

source .env && "$SLICER_PATH" --python-script scripts/run_optimization.py \
    SegmentEditorAdaptiveBrushTester/configs/default.yaml --trials 20 --no-exit

Resume previous study

bash

source .env && "$SLICER_PATH" --python-script scripts/run_optimization.py \
    SegmentEditorAdaptiveBrushTester/configs/tumor_optimization.yaml --resume

Tips

•Start with quick_test.yaml to verify setup
•Use --no-exit to inspect results visually in Slicer
•The SQLite database allows resuming interrupted runs
•Parameter importance helps focus future optimization
•Look for diminishing returns in trial history