Your knee signals trouble before pain does

How do you know if your knee is quietly getting worse — before it starts hurting?

For most people, the answer arrives too late: they seek help only once pain has become hard to ignore. Yet the musculoskeletal system rarely deteriorates without warning. Long before any discomfort registers, the body begins to compensate — shifting load, altering timing, borrowing stability from neighbouring structures. These adjustments happen below the threshold of conscious awareness, but they leave traces that are surprisingly readable once you know where to look.

Consider a few everyday examples. Uneven wear on the inner or outer edge of one shoe. A slight stiffness down one leg in the morning that eases within minutes. Needing to rock forward or use an armrest to get out of a chair — adding momentum to do what a healthy knee should manage alone. None of these feel dramatic. Each one is a quiet mechanical signal.

Research confirms that measurable gait changes — asymmetric stance time, a reduced stride length, a stiffened swing phase — precede both pain onset and the structural damage visible on X-ray. A 2024 clinical study of 34 patients with knee osteoarthritis found significant gait asymmetry before surgery: stance and double-limb support phases were disproportionately longer on the unaffected side, while the swing phase on the affected side was extended (p=0.004). Crucially, these deviations resolved after joint replacement, confirming they were functionally meaningful compensation patterns — not incidental findings.

The biomechanical signals gait analysis looks for

Several distinct patterns characterise a knee under biomechanical stress. Elevated knee adduction moment — the inward-twisting force generated at early stance — loads the medial compartment disproportionately, long before cartilage changes appear on imaging. Stiff-knee gait, in which the joint fails to flex fully during the swing phase, emerges as the body dampens movement to protect a compromised structure. These patterns reflect the same compensatory logic described in the preceding section, but they require objective capture to quantify reliably.

What clinical trials using MAI Motion have established is greater precision about which computational metrics carry the most diagnostic weight. Statistically significant biomarkers for knee pain (p<0.05) include the smoothness of the maximum knee flexion curve during a squat — how fluid, rather than hesitant or interrupted, that bending arc is — and, during sit-to-stand, cumulative acceleration across elbow and knee keypoints alongside an overall smoothness measure. These are not signals a clinician could reliably judge by eye; they are computational signatures derived from high-resolution frame-by-frame capture and only become legible once the movement data is processed at scale.

Reassuringly, the protocol is far from onerous. Research confirms that just three repetitions of a sit-to-stand test are sufficient to produce accurate knee joint measurements — brief enough for routine clinic appointments and practical for home re-scanning via the MAI Motion app.

Crucially, these markers respond to treatment. When an intervention is effective, the flexion curve becomes smoother and acceleration patterns improve measurably — a shift detectable between pre- and post-treatment scans. That reversibility is what distinguishes genuine functional biomarkers from background variation, and it is what makes objective monitoring of treatment efficacy possible.

How MAI Motion® turns a smartphone video into clinical data

The technology behind all of this sits in a standard smartphone. MAI Motion® — a UKCA and MHRA-registered AI motion analysis platform developed under Prof Paul Lee — requires no wearable sensors, no reflective markers, and no laboratory calibration. The patient simply moves; the system watches.

During a session, the smartphone camera tracks 15 skeletal keypoints simultaneously at 120 frames per second. That raw stream of 2D video is converted in real time into a 3D volumetric mesh — a spatial reconstruction of joint position and movement that goes considerably beyond what a flat image can show. The process generates thousands of data points per second of movement, capturing the precise angles, timing, and loading patterns that define how a knee is actually functioning.

The outputs are not impressions or estimates; they are reproducible, quantified measurements. Because the capture pipeline uses the same algorithm whether the camera is in a clinic or a patient's living room, follow-up monitoring scans can be completed at home via the app — with results feeding directly into the same biomechanical profile established at baseline. That continuity is deliberate: data gathered at each re-scan is genuinely comparable to the last, free from the variation that comes with changing equipment or environment.

MAI Motion® is deployed at MSK Doctors and, for patients based in London, through the London Cartilage Clinic at Harley Street.

What the C.R.A.F.T. framework assesses

Raw data — however rich — is clinically useful only once it has been interpreted. Within MAI Motion®, that interpretive layer is provided by the C.R.A.F.T. framework: a diagnostic methodology that evaluates knee function across four domains throughout the gait and movement cycle. These cover how the joint aligns under load, how it tracks through its arc of motion, the forces being transmitted at each phase, and the kinematics — the precise timing and geometry — of each movement pattern. Together, they translate thousands of per-second data points into a structured biomechanical profile of the knee.

The patient-facing output of that profile is the Motion Age score: a single figure expressing functional biological age, derived from movement alone and benchmarked against age-matched population norms. A 58-year-old whose knee mechanics resemble those of a healthy 44-year-old will see that reflected in their score; so will one whose movement patterns track older than their years. The number is concrete, comparable, and — critically — trackable.

Because the same capture pipeline runs at each clinic visit and at home via the app, Motion Age is updated longitudinally with every scan. This transforms the assessment from a one-time snapshot into a trajectory: patients and clinicians can observe whether function is improving, holding steady, or declining — and whether a given intervention is shifting the curve. Published clinical evidence indicates that most people following a structured programme see Motion Age fall meaningfully below their chronological age within 16 weeks, though individual outcomes depend on the nature and stage of the underlying condition.

Accuracy, limitations, and what smartphone capture can and cannot see

The strongest independent validation to date comes from a 2026 peer-reviewed PMC study testing a single smartphone camera with the MediaPipe pose-estimation framework against the OPAL wearable sensor — the accepted field gold standard — in 27 healthy volunteers. Mean absolute error for knee flexion reached 4.10°±2.32° on the right and 3.15°±3.10° on the left, with bilateral correlation of r=0.916 and r=0.845 respectively. For routine clinical monitoring of knee flexion, those figures sit within clinically acceptable bounds.

The broader picture across markerless systems is consistent: errors of 2°–6° versus VICON laboratory capture are typical, and agreement is strongest in the sagittal plane — the flexion-extension arc that dominates most knee assessments. Where accuracy falls off is in the frontal and transverse planes. Valgus and varus angles, and rotational metrics, remain materially weaker across markerless systems and should be interpreted accordingly.

MAI Motion®'s conversion of 2D video into a 3D volumetric mesh addresses some of the depth-estimation constraints inherent to flat-plane capture, and may improve frontal-plane performance over simpler pipelines. Whether it does so to a clinically significant degree remains unconfirmed — no published head-to-head comparison between MAI Motion® and VICON-grade laboratory capture exists. Separately, longitudinal trials comparing MAI Motion® gait biomarkers directly against clinical imaging for OA onset prediction have not been published. Both represent genuine evidence gaps, and are relevant context for clinicians reviewing the platform's outputs.

Getting assessed at MSK Doctors

Arranging a MAI Motion® assessment at MSK Doctors requires no GP referral and no waiting-list appointment. Patients in Lincolnshire and the wider East Midlands catchment are seen at the group's clinics in Sleaford (NG34) and Grantham (NG31); follow-up monitoring uses the same capture pipeline via the MAI Motion® app, without needing to return to clinic for each check.

The data does not interpret itself. Motion Age scores and C.R.A.F.T. outputs are reviewed with an MSK Doctors consultant, who places them in the context of your clinical history, symptoms, and any imaging — determining whether the findings support a monitoring plan, a targeted intervention, or further investigation. For patients based in London, the same technology is available through the London Cartilage Clinic at Harley Street.

The wider implication runs through everything covered here: how you move is a record of how your joints are managing load, and that record is now readable earlier, more precisely, and with far less disruption than clinical practice has previously allowed. Whether that changes anything for an individual depends on a consultant assessment — but the window for acting before pain forces the issue is larger than most people realise.

To arrange an assessment, visit mskdoctors.com to book directly.