Seattle Spine Align

Genetic-algorithm surgical plan optimizer · research preview

Patient parameters

A deterministic real cohort case (first patient in the cohort) is auto-loaded on page open so local and live deployments start from the same patient for side-by-side comparison. Click Find top 5 plans to run the genetic algorithm against this patient. Use Random real patient to pick a different cohort case (with precomputed top-5 plans for Forward mode), or Random simulated patient to generate a synthetic deformity for a live GA run.

Spinopelvic (required)
Evaluation mode
Choose the forward model + decision space the planner uses:
Hybrid model uses the full literature maneuver library (ALIF, ACR, XLIF, OLIF, TLIF, PLIF, PSO, SPO, VCR with per-level degrees of correction). Forward model uses the 281-patient cohort's surgical maneuvers only (ALIF, XLIF, TLIF, and PSO) and does not show the per-level degrees of lordosis.
Alignment goals
Check every goal that should drive the GA. Combinations stack: e.g. GAP + T4-L1PA weights both. Custom adds your surgeon-entered constraints (table below) on top. Surgeons who do not subscribe to a given framework should leave that checkbox unchecked.
Targets are enforced at the timepoint: and max acceptable mech-fail probability (%) .
Genetic-algorithm generations
Penalty tuning (click rows to enable/disable or override weights)
Show / hide · default values shown · changes apply to next GA run only

Alignment-target soft penalties (Term 1)

Each row is one constraint. Toggle to include or skip; weight is the per-constraint multiplier.

Use Penalty Weight Reference

Mechanical-failure proxy factors (Term 2)

Each factor is a literature-derived multiplier. Toggle to include or skip; magnitudes are not exposed because they come from cited papers.

Use Factor Reference

Anatomic-risk surcharges (Terms 3 and 4)

Conus-anatomy fitness surcharges. Toggle to include or skip; magnitude is the flat fitness-unit add.

Use Surcharge Magnitude Reference

Note: The runtime scales with the number of generations selected. At 6 generations (the default) the run typically finishes in 40-60 seconds; lower values complete faster, up to 8 takes 60-90 seconds.

Preoperative spine (live)

Updates as you change PI, LL, PT, L4-S1, TK.