Dana, 38 · runner

Movement

Right stanceMeasured

0^°

Knee frontal-plane angle

Right stanceModeled

Patellofemoral stress index

Week 0 · initial evaluation

Wk 0 · eval2468 · re-test

Month 7 · each criterion ≥ 90%

Your clinic's return-to-sport criteria: 3 of 5 met.

✓ 7 moSince surgery✓ 94%Hamstring✓ 94/92%Timed, triple✗ 83%Quadriceps ✗ 80%1-leg jump90/89%Single, cross66°/78°Landing R/L41%R force · model

Assessment report · draft

Week 0 · initial evaluation

For clinician review · unsigned

Findings · left vs rightMeasuredModeled

What the cameras saw

Measured rows come from 3D joint centres; small angles show a ±4.5° range. MDC is the smallest change that is more than measurement noise (approximate, from published test-retest work); your clinic's own repeat measures replace it. Modeled rows come from a musculoskeletal model driven by those angles and are shown as a relative index only. Reference values are published group means (Herrington; Bramah 2019) or estimates, not cut-offs.

What this suggestsHypotheses

A hip and ankle story, not a knee story

For clinician reviewDecision support only. Confirm with hip strength testing, ankle and hip range of motion, and symptom response.

Plan of care · draftClinician edits

Four progressions, re-measured every 2 weeks

How it is captured

Two synchronized cameras about 3.5 m away, 30–45° either side of the patient, 60 fps. Their 2D keypoints are triangulated into 3D joint centres.
Accuracy, approximately: joint angles land within about 4.5° of a marker-based motion lab in the published OpenCap validation (Uhlrich 2023). Frontal-plane and rotation angles are the least reliable, so small angles are shown as ranges.
Modeled layer: ground force, joint moments and a patellofemoral stress index come from a musculoskeletal simulation (OpenSim-style) driven by the measured angles. In published OpenCap work, ground force errs by about 6% of body weight on average and joint moments by about 1.2% of body weight × height, validated for walking, squats, sit-to-stand and drop jumps, not running. Muscle demand is shown only as low, moderate or high: model-estimated muscle activity agrees poorly with EMG. Ligaments are drawn for anatomy reference; no strain is estimated.
One ordinary camera is not enough for frontal-plane knee angles in 3D. A single depth camera is the minimum.

Body Map

A movement lab from two calibrated cameras. Each movement is rebuilt as a 3D skeleton: joint angles are measured, and forces and joint loads are modeled, for clinician review.

What you're seeing

A simulated patient, not a recording. Dana's right knee drifts inward, the opposite side of the pelvis drops and the right ankle bends less. Switch layers to compare what is measured (angles, timing) with what is modeled (ground force, joint load, muscle demand). Drag to orbit. Scrub 8 weeks of care to watch the pattern settle, or try a higher running cadence. The second case, Marcus, shows why passing hop tests is not the whole story after ACL reconstruction.

How it works

Two synced cameras give 3D joint centres. From those, Room computes the angles physical therapists already use: frontal-plane knee angle, pelvic drop, trunk lean, hip adduction and ankle dorsiflexion. A musculoskeletal model, the kind OpenSim and OpenCap use, then estimates ground force and joint moments. The trustworthy outputs are left-right differences and change over time.

In a real clinic

The same few minutes of movement at the first visit and every re-test, about 6 minutes of a 45-minute evaluation. The clinician gets numbers they would otherwise eyeball, draft findings and a progress chart the patient can understand. They decide what it means.

Every measure carries its minimal detectable change (MDC), and a change smaller than the MDC is not called progress. Patient-reported outcomes the clinic already uses sit beside the movement numbers: LEFS, KOOS, QuickDASH for the upper limb, and PSFS. The hop and jump battery is a physical performance test, which clinics bill as CPT 97750 where the payer covers it.

What it is not

Room measures movement. It does not see cartilage, tendons or ligaments, and it does not decide what is wrong. Anatomy here is drawn for reference. Modeled values carry wider error than measured angles, and every finding is decision support for the clinician. Return-to-sport criteria are the clinic's own; Room only fills in the numbers. In development and not FDA-cleared.