How Rectify Measures Your Spine in 3D

You buy a wearable. It promises to track your posture. Back health finally measurable — that's what it says on the box.

You wear it. It vibrates when you hunch. Shows a curve in the app. Feels precise.

It isn't. Not for the spine.

This isn't about build quality. Not the app. It's about the physical measurement principle — and that has a problem no software update can fix.

Back pain (M54) was the most common single diagnosis for sick leave in Germany in 2024. At KKH, this diagnosis caused 953,000 lost workdays and affected one in eight employed people — 20 percent more than in 2019 [10]. The RKI BURDEN project records a 12-month prevalence of 61.3 percent in the general population [9]. More than six in ten adults know back pain from personal experience.

This article explains why the standard measurement principle is structurally unsuitable for the spine — and how a different principle solves the problem.

What an IMU Actually Measures — and What It Doesn't

IMU stands for Inertial Measurement Unit. A classic IMU contains three sensors: accelerometer, gyroscope, magnetometer. Together they deliver six to nine measurements per time step.

The gyroscope measures angular velocity — how fast a body rotates, in degrees per second. What it does not measure: the current angle itself.

That's where the problem lies.

To calculate an angle θ(t) from angular velocity ω(t), you have to integrate:

θ(t) = ∫ω(t) dt

Each integration interval dt is short — typically 4 to 10 milliseconds. But every step contains a small measurement error ε: sensor noise, temperature effects, mechanical vibration. These errors add up. Step by step. The result: accumulated drift.

Al Borno and colleagues wrote in 2022 in the Journal of NeuroEngineering and Rehabilitation: "drift, or the accumulation of error over time, inhibits the accurate measurement of movement over long durations." [2]

Simpson, Maharaj, and Mobbs stated even more fundamentally in 2019: "The absolute nature of gyroscopes raises an issue of bias error due to drift; longer duration validation studies are required." [1]

How large is this error in practice? Validation data from the CDC repository show RMSD orientation errors of 4.1 to 6.6 degrees for off-the-shelf IMU units during spinal measurement [4]. Haddas and colleagues achieved in 2026 under laboratory conditions — with specific calibration and subject-specific drift correction — an RMSD of just under 5 degrees for a single T1 sensor (ICC 0.89–0.96) [5]. That is the upper limit of what current technology delivers under optimal conditions. In everyday use, it's higher.

For spinal diagnostics, this is critical. The difference between physiological lumbar lordosis and early-onset rounded back lies in the range of 2 to 4 degrees. A sensor with a 5-degree error doesn't measure there — it guesses.

The Location Error: When a Sensor Doesn't Know Where the Curve Begins

An IMU on the chest measures a forward tilt of 45 degrees. What happened?

Option A: You hinged at the hip, lumbar spine remains in neutral lordosis. Hip hinge. The spine is relaxed, discs evenly loaded.

Option B: The hip stays extended, the lumbar spine flexes into hyperkyphosis. All 45 degrees come from the back. Intradiscal pressure rises steeply, disc fibers carry load asymmetrically.

The sensor sees both as an identical signal. 45 degrees. Done.

This isn't a calibration problem. A single sensor measures the global rotation of its attachment point. It cannot see where that rotation originates.

Moon and colleagues studied this directly in 2022. Finding: to separate pelvic movement from lumbar flexion, at least three IMUs are needed — one on the pelvis, one in the lumbar region, one on the thorax. And even then, drift remained "a significant problem" in their 3-IMU cluster system [3].

The underlying concept is lumbopelvic rhythm. When bending forward, flexion distributes between the hip joint and lumbar spine in a characteristic ratio. This ratio varies between individuals, changes with pain and fatigue — and directly determines where mechanical load acts on the discs.

A single sensor doesn't know this rhythm. It only knows its own tilt in the world.

Drift and Magnetic Interference: Why Software Doesn't Solve the Problem

Modern IMU systems fight drift with fusion algorithms. Madgwick filter, Kalman filter, VQF — all fuse the three sensor channels to compensate errors mutually.

The magnetometer serves as an absolute reference: it knows where north is. This stabilizes the orientation estimate.

Works outdoors. Less so in the office.

Fan, Li, and Liu investigated in 2017 how magnetic disturbances affect orientation estimation [6]. Indoor environments — metal desks, PCs, monitors, steel chairs, reinforced concrete ceilings — generate local interference fields. The magnetometer no longer reads a homogeneous earth field, but a superposition of earth and device fields. Heading errors above 10 degrees in such environments without active compensation are the norm.

The typical office workstation is a worst-case scenario for a magnetometer-based wearable.

The factory floor is even worse: metal shelving, conveyor systems, electric motors. Exactly the environments where ergonomic monitoring is most urgently needed.

Fusion algorithms reduce the problem. They don't solve it. Al Borno and colleagues confirmed this in 2022: IMU fusion compensates drift, it doesn't eliminate it [2]. The mathematics do not allow reconstruction of an accumulated integration error — because the information about the true path is gone.

Direct Deformation Measurement: The Principle Behind the FlexTail

There is a different approach. One where no integral is formed.

A strain gauge measures electrical resistance. And resistance is not an accumulated value. It is a state. Right now.

The principle: the resistance of an electrical conductor is R = ρL/A — conductivity ρ, length L, cross-sectional area A. When a strain gauge is stretched, L and A change simultaneously. The resistance changes proportionally to the mechanical strain ε. This ratio is called the gauge factor:

GF = (ΔR/R) / ε

For classic metal foil strain gauges, GF is 2.0 to 2.2. Printed silver nanoparticle gauges on PET film achieve GF values between 1.74 and 2.03 — sufficient for spinal bending measurement [8].

PET stands for polyethylene terephthalate. Flexible, chemically stable, compatible with silver inks for printing. As a substrate for printed strain gauges, it is validated: linear characteristics across bending angles from 0 to 90 degrees [8].

The key point: the sensor measures what exists right now. No time integral. No accumulated error. When you straighten up, it shows zero. When you bend, it shows the bend.

The FlexTail contains 36 sensor pairs along its length — 18 measurement points from sacrum to upper thoracic spine. Each pair measures the local curvature at its location.

From these 18 local curvature measurements, the complete spinal shape in 3D can be reconstructed. The algorithmic method is based on Euler-Bernoulli beam theory: from local curvature κ and orientation α at each point, the Cartesian 3D coordinates of the bending line are calculated by integration.

Francoeur and colleagues validated this principle at IEEE ICRA 2024. Reconstructing needle shapes, they achieved an RMSE of 0.58 to 0.66 millimeters [7]. The method transfers directly to distributed strain gauge arrays.

A single IMU at the sacrum serves as a spatial anchor: it anchors the relatively measured spinal shape in the world coordinate system. For this task, one IMU is enough. The sacrum moves slowly in physiological daily use, with no long phases of continuous rotation. Drift stays within clinically tolerable limits — because no long-term angular integration is needed for the shape reconstruction itself.

The result: a complete, segmental 3D map of the spine. Not a tilt. A shape.

What This Means in Practice

IMUs are not bad sensors. They have a well-defined application range.

Short-duration measurements — gait analysis, jump force, arm rotation in sports — work well. Drift doesn't have time to build up.

Global posture trends over minutes can be captured usefully. Is someone sitting straight or slouched? Roughly detectable, when the environment is magnetically clean.

What IMUs structurally cannot do: segmental spinal measurement over hours, in magnetically disturbed environments, without at least three sensors for pelvis-spine differentiation, without laboratory-grade calibration.

This has consequences:

Ergonomics engineers in industrial settings need reliable data on cumulative spinal load over full shifts. A wearable with 5-degree drift after 45 minutes on the factory floor — plus magnetic field interference from the conveyor — provides no usable basis for risk assessments.

Physiotherapists in spinal rehabilitation want to know: is the patient applying the learned movement patterns at home? Are they using the hip hinge when lifting? A single sensor cannot answer that. A sensor that sees the entire spinal curve can.

Clinical research on scoliosis, kyphosis, and disc degeneration requires accurate longitudinal data over weeks and months. Segmental 3D reconstruction is not optional here — it is the measurement variable itself.

Conclusion

The problem isn't sensor quality. Drift is not a firmware weakness. It's mathematics.

Anyone who wants to calculate an angle from angular velocity must integrate. And integration accumulates error. In the spine, where clinically relevant differences lie in the 2 to 4 degree range, that's enough to render the measurement useless.

Strain gauges circumvent this structurally. They don't measure an integrated state, but the current one. Direct, drift-free, segmental.

Anyone who wants to measure the spine across a workday needs a method that measures locally — not one that integrates globally.

Sources

[1] Simpson LA, Maharaj A, Mobbs RJ. The role of wearables for orthopedic interventions: systematic review and meta-analysis. BMC Musculoskeletal Disorders. 2019. DOI: 10.1186/s12891-019-2430-6

[2] Al Borno M, Uhlrich SD, Hicks JL, et al. OpenSense: Validating IMU-Based Movement Analysis During Varied Activities. Journal of NeuroEngineering and Rehabilitation. 2022. DOI: 10.1186/s12984-022-01001-x

[3] Moon Y, Ozturk O, et al. Extraction of Lumbar Spine Motion Using a 3-IMU Wearable Cluster. Sensors (MDPI). 2022. PMC9823955

[4] Schall MC, Fethke NB, Chen H, et al. Accuracy and repeatability of an inertial measurement unit system for field-based occupational studies. Ergonomics. 2016;59(4):591–602. https://stacks.cdc.gov/view/cdc/121211

[5] Haddas R, et al. Translating biomechanics to clinic: validating a spine-specific wearable for remote functional assessment. Spine Journal. 2026. DOI: 10.1016/j.spinee.2026.01.025

[6] Fan B, Li Q, Liu T. How Magnetic Disturbance Influences the Attitude and Heading in Magnetic and Inertial Sensor-Based Orientation Estimation. Sensors (MDPI). 2017. DOI: 10.3390/s18010076

[7] Francoeur J, et al. Shape Reconstruction from Distributed Strain Measurements Using Euler-Bernoulli Beam Theory. IEEE ICRA 2024. PMC11507468

[8] NovaCentrix Engineering Team. Gauge Factor of Printed Strain Gauges on PET and Kapton Substrates. NovaCentrix Technical Whitepaper. https://www.novacentrix.com

[9] Robert Koch Institut. BURDEN 2020 — Back Pain Prevalence Germany. 2020.

[10] KKH Kaufmännische Krankenkasse. Back pain 2024 main reason for sick leave at work. Press release, 24 November 2025. https://www.kkh.de/presse/pressemeldungen/haeufigstediagnosenjob