The Science

What Stela measures, and the research behind it.

Balance is one of the few health metrics that moves before its consequences. Stela measures it.

Stela measures postural sway using the inertial sensors already in your phone. The measurement is drawn from a class of instruments validated across more than two decades of independent research. This page explains what we measure, how we measure it, and the evidence we rely on.

What is measured

When you stand still, your body is never truly still. Small continuous adjustments keep you upright. The velocity and variability of those adjustments, called postural sway, reflect how the neuromuscular, vestibular, and visual systems work together. Sway changes with age, with injury, with medication, and with conditions that affect the nervous system.

Stela's primary metric is acceleration path rate, measured in meters per second cubed (m/s³). It is the cumulative movement of the phone through acceleration space during a stance, divided by the duration of that stance. It belongs to the family of accelerometry-based sway metrics established in trunk accelerometry research since the late 1990s [1, 2, 3]. It is dimensionally distinct from force-plate center-of-pressure velocity (m/s) and is internally consistent for normative comparison within the accelerometry class. The metric is not proprietary.

Test-retest reliability

Reliability describes whether the same person, tested twice under the same conditions, gets the same result. It is reported as an intraclass correlation coefficient (ICC), where 1.0 is perfect and values above 0.75 are generally considered good to excellent.

StudyPlatformICC Range
Saunders et al. 2015 [4]Lower back accelerometer0.736–0.972 (trial-to-trial); 0.760–0.954 (block-to-block)
Prato et al. 2024 [5]Android and iOS smartphones, lower-back beltEO: 0.82–0.98 (iOS), 0.84–0.96 (Android). EC: 0.34–0.79 (iOS), 0.41–0.87 (Android)
Moe-Nilssen 1998 [2]Triaxial accelerometer, lumbar spineStanding two-foot: ICC(3,1) < 0.56 (low relative reliability, restricted range; high absolute repeatability). Standing one-foot, mediolateral: 0.84. Walking tests: 0.79–0.94
Mancini et al. 2012 [1]Body-worn accelerometer at L5JERK (time derivative of acceleration): 0.86 (PD), 0.87 (CTR). Time-domain measures (RMS, mean velocity, etc.): 0.55–0.84 (PD), 0.60–0.89 (CTR)
Alqahtani et al. 2020 [3]Lower back accelerometerNormalized path length (NPL): 0.49–0.82. RMS: 0.41–0.83
Calcagni et al. 2025 [19]Smartphone (NeuFun-TS)0.71 (eyes-closed feet-together stance)

JERK in Mancini et al. 2012 is the time derivative of acceleration in units of m/s³, which is the same dimensional quantity as Stela's acceleration path rate. Liparoti et al. 2021 [20] reviewed 19 studies of IMU-based sway reliability and reported moderate to excellent test-retest values across static balance conditions, with lower-back placement consistently showing the strongest reliability. Absolute numeric values differ across studies due to filter design, sample rate, hardware, and the specific sway parameter reported; the shared construct is cumulative path in acceleration space divided by duration. Stela-specific reliability data for the full 8-test battery will be reported.

The assessment battery

Stela administers an 8-test core battery organized into four categories. Each stance is timed; all stances are 30 seconds.

CategoryTestWhat it isolates
Static
Body still, surface firm. Establish your baseline.		Bipedal Stance, Eyes Open (BSEO)All three balance systems active: visual, vestibular, and proprioceptive. Establishes the minimum-sway reference every other test is compared to.
Bipedal Stance, Eyes Closed (BSEC)Vision removed. Vestibular and proprioceptive systems carry the full load. The Romberg ratio (sway here vs. eyes-open) quantifies visual dependency.
Dynamic
Base of support or sensory input manipulated.		Narrow Stance, Eyes Closed (NSEC)Narrowed base of support with vision removed, increasing the proprioceptive demand relative to BSEC.
Head Turns, Eyes Open (HTR-EO)Active VOR engagement via paced horizontal head rotation at 0.75 Hz, approximately 30° each side, while fixating on a visual target about 1 m away at eye level.
Reactive
Unexpected demands placed on a stable stance.		Unexpected Eye Closure (UOC)Reactive postural control. An auditory cue fires at a random time between 3 and 15 seconds into the stance, prompting the user to close their eyes. The sway response in the 3-second window after the cue reflects a distinct neural pathway from anticipatory balance adjustment.
Forward Head Pitch (FHP)Sustained VOR suppression under forward head pitch (approximately 30°, chin toward chest). The sway delta vs. the BSEO baseline is the primary signal. Inability to suppress the VOR in this position has been associated with reduced postural stability.
Dual-Task
Stand and think simultaneously. Key metric is dual-task cost.		Serial Subtraction (DT-A)Cognitive-motor interference via prefrontal arithmetic load. User counts backward aloud by 3 from a randomized starting number while maintaining bipedal stance, eyes open. Dual-task cost (sway increase over BSEO baseline) quantifies the attentional demands of balance.
Verbal Fluency (DT-V)Semantic memory retrieval under postural load on a narrower base (semi-tandem stance, eyes open). User names animals aloud continuously. A different executive pathway than arithmetic; together they map the dual-task cost surface.

Opt-out is available on every test. An inability to complete a test is recorded as a scored outcome (score of 0, opt-out flagged) rather than a skipped test; missingness is treated as signal. Individual stance conditions have established precedent in the literature. The battery combination, sequencing, and smartphone-based delivery of reactive and dual-task conditions are Stela-defined. Reliability data for all eight conditions in Stela's implementation will be reported.

Placement and hardware

Stela supports two placement tracks. In the clinical track, a coordinator positions the phone against the L4/L5 lumbar region using a standardized sleeve. This approximates the body's center of mass and is consistent with published practice for trunk sway measurement [4, 18]. In the consumer track, the user holds the phone steady and flat for the duration of each stance, following on-screen guidance. Each track yields different absolute readings but maintains consistent longitudinal reference within itself.

What Stela is and is not

Stela is a measurement and data platform. It produces quantitative postural sway data intended for self-observation, clinical supplementation, and longitudinal tracking. It is not a diagnostic device. It does not screen for, diagnose, or treat any medical condition. Clinical judgment remains with the supervising clinician.

Data handling

Identifying information is encrypted at rest using field-level KMS encryption. De-identified data used for population reference curves is stored separately from identified data by architectural design, not policy alone. Each session contributes data, improving the measurement for everyone. Full detail is in our privacy summary.

References

  1. Mancini M, Salarian A, Carlson-Kuhta P, et al. (2012). ISway: a sensitive, valid and reliable measure of postural control. J NeuroEng Rehabil, 9(1), 59.
  2. Moe-Nilssen R. (1998). Test-retest reliability of trunk accelerometry during standing and walking. Arch Phys Med Rehabil, 79(11), 1377–1385.
  3. Alqahtani BA, Sparto PJ, Whitney SL, et al. (2020). Psychometric properties of instrumented postural sway measures in independent living older adults. BMC Geriatrics, 20(1), 82.
  4. Saunders NW, Koutakis P, Kloos AD, et al. (2015). Reliability and validity of a wireless accelerometer for the assessment of postural sway. J Appl Biomech, 31(3), 159–163.
  5. Prato TA, Lynall RC, Howell DR. (2024). Validity and reliability of an integrated smartphone measurement approach for balance. J Sport Rehabil, 34(2), 177–184.
  6. Anthony EC, Kam OK, Klisch SM, et al. (2024). Balance assessment using a handheld smartphone with principal component analysis for anatomical calibration. Sensors, 24(17), 5467.
  7. Kavanagh JJ, Menz HB. (2008). Accelerometry: a technique for quantifying movement patterns during walking. Gait & Posture, 28(1), 1–15.
  8. Van der Kooij H, Campbell AD, Carpenter MG. (2011). Sampling duration effects on centre of pressure descriptive measures. Gait & Posture, 34(1), 19–24.
  9. Prieto TE, Myklebust JB, Hoffmann RG, et al. (1996). Measures of postural steadiness: differences between healthy young and elderly adults. IEEE Trans Biomed Eng, 43(9), 956–966.
  10. Mahony R, Hamel T, Pflimlin JM. (2008). Nonlinear complementary filters on the special orthogonal group. IEEE Trans Autom Control, 53(5), 1203–1218.
  11. Madgwick S, Harrison A, Vaidyanathan R. (2011). Estimation of IMU and MARG orientation using a gradient descent algorithm. IEEE Int Conf Rehabil Robot.
  12. Luinge HJ, Veltink PH. (2005). Measuring orientation of human body segments using miniature gyroscopes and accelerometers. Med Biol Eng Comput, 43(2), 273–282.
  13. Winter DA, Patla AE, Prince F, et al. (1998). Stiffness control of balance in quiet standing. J Neurophysiol, 80(3), 1211–1221.
  14. Peterka RJ. (2002). Sensorimotor integration in human postural control. J Neurophysiol, 88(3), 1097–1118.
  15. Mancini M, Horak FB. (2010). The relevance of clinical balance assessment tools to differentiate balance deficits. Eur J Phys Rehabil Med, 46(2), 239–248.
  16. Lo PY, Su BL, You YL, et al. (2022). Measuring the reliability of postural sway measurements for a static standing task: the effect of age. Front Physiol, 13, 850707.
  17. Patel S, Park H, Bonato P, et al. (2012). A review of wearable sensors and systems with application in rehabilitation. J NeuroEng Rehabil, 9(1), 21.
  18. Mancini M, Horak FB, Zampieri C, et al. (2011). Trunk accelerometry reveals postural instability in untreated Parkinson's disease. Parkinsonism Relat Disord, 17(7), 557–562.
  19. Calcagni M, Kosa P, Bielekova B. (2025). Smartphone postural sway and pronator drift tests as measures of neurological disability. BMC Neurology, 25(1), 50.
  20. Liparoti M, et al. (2021). Postural stability assessment during quiet standing: reliability of instrumented measurement in clinical practice. Systematic review; 19 studies. PMC8348903.
  21. Yin L, Xu X, Wang R, et al. (2025). Validity and reliability of inertial measurement units on gait, static balance and functional mobility performance among community-dwelling older adults: systematic review and meta-analysis. EFORT Open Reviews, 10(4), 172–185.
  22. Cole TJ, Green PJ. (1992). Smoothing reference centile curves: the LMS method and penalized likelihood. Stat Med, 11(10), 1305–1319.
  23. WHO Multicentre Growth Reference Study Group. (2006). WHO Child Growth Standards. World Health Organization.
  24. Raymakers JA, Samson MM, Verhaar HJ. (2005). The assessment of body sway and the choice of the stability parameter(s). Gait & Posture, 21(1), 48–58.