Background and Rationale Quantitative gait analysis is an established method for assessing human movement, lower-limb function, mobility, and biomechanical performance. The current technical gold standard for precise gait analysis is optical motion capture, such as Vicon-based three-dimensional motion analysis. These systems provide high-resolution spatiotemporal and kinematic data, but their use is largely restricted to specialized gait laboratories because they require dedicated infrastructure, trained personnel, time-consuming preparation, and substantial financial resources. Consequently, gold-standard gait analysis is rarely available in routine clinical care, large-scale research settings, or home-based longitudinal monitoring.
In contrast, smartphones and wearable devices contain inertial measurement units, including accelerometers, gyroscopes, and magnetometers, which allow the recording of movement-related raw sensor data in a highly scalable and low-threshold manner. AI-based algorithms may enable automated extraction of clinically and biomechanically relevant gait parameters from such sensor data without the need for complex laboratory equipment. This creates the prospect of continuous or repeated gait assessment in clinical, ambulatory, or home-based environments.
However, before AI-based smartphone gait analysis can be interpreted as a valid technical alternative or complement to established laboratory-based systems, its agreement with the gold standard must be quantified under controlled experimental conditions. Key questions include whether smartphone-derived gait parameters correlate sufficiently with motion-capture-derived parameters, whether systematic measurement bias exists, how large the measurement error is, whether repeated measurements are reliable, and whether device platform or clothing-related factors influence measurement accuracy.
This study therefore investigates the technical comparability of AI-based smartphone-derived gait parameters with simultaneous Vicon optical motion-capture measurements in healthy adult volunteers. By quantifying agreement, measurement deviation, test-retest reliability, and potential sources of bias, the study aims to provide a technical foundation for future clinical and scientific use of smartphone-based gait assessment.
Study Objective The primary objective of this study is to evaluate the agreement between AI-based smartphone-derived gait parameters and gait parameters obtained from an optical Vicon motion-capture system.
The central research question is: To what extent do AI-based smartphone-derived gait parameters agree with measurements obtained using an optical Vicon motion-capture system?
The study will quantify the agreement between both measurement methods using established statistical measures, including intraclass correlation coefficients, Bland-Altman analysis, mean absolute error, and root mean square error.
Primary Hypothesis The AI-based smartphone-derived gait parameters show good agreement with Vicon-based gait analysis.
The formal hypothesis for the primary agreement analysis is:
H0: ICC ≤ 0.75, indicating insufficient agreement. H1: ICC > 0.75, indicating good agreement.
Study Design This is a monocentric, prospective cohort study conducted under controlled laboratory conditions. The study is exploratory in nature. It is designed as a technical validation study comparing smartphone-derived gait parameters with simultaneous measurements from an optical motion-capture reference system. No therapeutic intervention is performed, and no clinical treatment decisions will be made based on the study measurements.
Study Population The study will include 40 healthy adult volunteers. Participants must be at least 18 years old and must not have acute or chronic disorders of the lower extremities. Participants must be able to walk independently and must be able to perform a short standardized walking test. Study information and participant support will be available in German, English, and Spanish. No vulnerable participants or vulnerable patient groups will be included.
Inclusion criteria
Adults aged 18 years or older. No known acute or chronic disorder or injury of the lower extremities. Ability to walk independently. Sufficient physical capacity to complete a standardized walking test. Ability to understand the study information and provide written informed consent.
German-, English-, or Spanish-speaking. Exclusion Criteria: Acute injury or disease of the lower extremity. Chronic injury or disease of the lower extremity. Inability to walk independently. Relevant motor impairment or other major limitation affecting safe gait testing. Lack of capacity to provide informed consent. Inability to understand German, English, or Spanish. Insufficient physical capacity to perform the walking test. Any condition that, in the opinion of the study team, would impair participant safety or the validity of the gait measurements.
Recruitment and Consent Process: Participants will be recruited through a news email sent to employees and students of the TUM School of Medicine and Health. Interested individuals may register for participation via the TUM study portal and select an available test appointment. After registration, participants will receive a confirmation email including the test date, participant information, informed consent documents, information about the study site, study procedures, clothing requirements, and contact information. Participants will be asked to review the study documents before the study visit and to bring the completed documents to the appointment.
On the day of testing, the study team will provide personal information and clarification in person before any study-specific procedure is performed. Written informed consent will be obtained before the start of testing. Participation is voluntary, and participants may withdraw consent at any time without providing a reason and without any disadvantage.
Study Setting:
Testing will take place at the TUM Campus in the Olympiapark. The study will be conducted in a controlled gait-analysis environment using simultaneous smartphone-based inertial sensor recording and Vicon optical motion capture.
Study Procedures:
After arrival and completion of the informed consent process, participants will be prepared for the gait analysis. Vicon markers will be attached according to the requirements of the motion-capture protocol. Participants will then receive study smartphones, including an iPhone and an Android device. The devices will be placed in the trouser pockets so that both smartphone-based measurements and Vicon motion-capture data can be acquired simultaneously.
The test procedure consists of repeated walking trials. Participants will first perform the walking test wearing their own street trousers. During this condition, an iPhone and an Android smartphone will be carried in the respective trouser pockets. Participants will then walk for two minutes on the Vicon gait-analysis setup while all device data and Vicon data are recorded in parallel. This procedure will be repeated three times. After completion of the first measurement condition, participants will change into standardized trousers with two pockets positioned at thigh level. The walking tests will then be repeated under this standardized clothing condition. Again, the participant will perform three two-minute walking trials while both smartphone and Vicon measurements are recorded simultaneously.
This design allows assessment of the agreement between smartphone-based and Vicon-based gait parameters, comparison between iOS and Android devices, assessment of repeated measurements, and evaluation of the influence of clothing and pocket position on measurement performance.
The total duration of participation, including arrival, preparation, consent confirmation, marker placement, testing, and completion of the visit, is expected to be approximately 90 minutes. The active testing time is expected to be approximately 45 minutes.
Devices and Measurement Systems:
The study uses smartphones and wearable devices equipped with inertial sensors, including accelerometers, gyroscopes, and magnetometers. The devices record raw movement data relevant to the study objectives.
The Study-App will be used to record and process raw inertial sensor data. The application evaluates raw accelerometer, gyroscope, and magnetometer data and presents derived results without clinical interpretation or medical decision-making. The devices and application are not used as certified medical devices for diagnostic or therapeutic purposes within this study. No CE certification for the specific study purpose is available or sought in the context of this technical validation study. The Vicon optical motion-capture system serves as the reference standard for gait analysis. Vicon-derived gait parameters will be used as the comparator for smartphone-derived parameters.
Data Sources
- Smartphone-based and wearable-based inertial sensor data.
- Vicon optical motion-capture data.
- Derived AI-based gait parameters calculated from inertial sensor data.
- Derived spatiotemporal and kinematic gait parameters from the Vicon reference system.
- No routine clinical data are required for the primary technical comparison.
Primary Outcome Measures The primary outcome measures assess agreement and measurement error between smartphone-derived gait parameters and Vicon-derived gait parameters.
Primary statistical outcome measures include: Intraclass correlation coefficient for central gait parameters, Bland-Altman limits of agreement, Mean absolute error, Root mean square error.
These agreement and error measures will be calculated for the following gait parameters: Gait speed, Step length, Cadence, Step time, Double support time, Gait asymmetry, Kinematic angle parameters of the lower extremity, Knee joint range of motion.
Secondary outcome measures include: Comparison of measurement performance between Android and iOS devices, Test-retest reliability across repeated walking trials, Analysis of systematic bias between smartphone-derived and Vicon-derived measurements, Influence of trouser type and pocket position, comparing participants' own trousers with standardized trousers; Assessment of intra-individual variability across repeated measurements; Exploratory evaluation of whether measurement deviations depend on walking speed or individual participant-related factors.
Statistical and Biometric Analysis:
The study is exploratory and focuses on technical agreement between measurement systems. Spatiotemporal and kinematic gait parameters will be obtained from smartphone-based inertial sensor data and from the Vicon optical motion-capture system.
For each relevant gait parameter, the agreement between smartphone-derived and Vicon-derived measurements will be quantified using intraclass correlation coefficients. The primary hypothesis will be evaluated using an ICC threshold of 0.75, with ICC values above 0.75 interpreted as indicating good agreement.
Bland-Altman analyses will be performed to assess bias and 95% limits of agreement between the two measurement methods. The mean difference between smartphone-derived and Vicon-derived values will be used to quantify systematic bias. Mean absolute error and root mean square error will be calculated to quantify absolute and squared deviations from the reference system.
Repeated measurements will be used to evaluate test-retest reliability and intra-individual variability. Secondary analyses will compare iOS and Android devices and examine the effect of clothing condition. The standardized trouser condition will help determine whether defined pocket position improves measurement consistency compared with participants' own trousers.
The planned sample size is 40 participants. As this is an exploratory technical validation study, the sample size is intended to provide sufficient repeated-measurement data to estimate agreement, measurement error, and variability across devices and test conditions.
Risk-Benefit Assessment This is a non-invasive study involving healthy adult volunteers. The study does not provide direct individual benefit to participants. However, it is expected to generate relevant scientific and technical knowledge regarding the validity, reliability, and limitations of AI-based smartphone gait analysis. This may support future development of scalable gait-assessment tools for research, clinical follow-up, rehabilitation, and longitudinal mobility monitoring.
The burden and risks associated with participation are minimal. The study consists of short walking tasks under controlled laboratory conditions, corresponding to ordinary physical activity. Potential risks are limited to mild fatigue, transient discomfort, a very low risk of stumbling, and rare minor skin irritation caused by motion-capture markers. Testing is performed in a supervised laboratory environment by trained personnel, thereby further minimizing risk.
No additional costs arise for participants. No incidental clinical findings are expected, as the study is designed as a technical measurement comparison and does not include diagnostic evaluation.
Overall, the expected scientific and potential future clinical benefit outweighs the minimal risk and burden associated with study participation.
Withdrawal and Discontinuation Participants may withdraw from the study at any time without giving a reason and without any disadvantage. Individual participation will be discontinued if a participant experiences pain, dizziness, insecurity while walking, or any newly occurring health limitation that may impair safe completion of the walking tasks. Participation may also be discontinued if the study team determines that participant safety or data quality can no longer be ensured. Technical problems that prevent valid data acquisition may also lead to discontinuation of the individual study visit. The entire study may be terminated early if the required number of participants has been reached or if technical defects of the measurement systems prevent valid continuation of data collection.
Data Protection and Data Management The legal basis for data processing is the participant's informed consent in accordance with General Data Protection Regulation. The responsible institution for data processing is TUM University Hospital. The study will not collect directly identifying personal data within the measurement dataset. Each participant will receive a randomly generated study ID. The gait data will be processed in anonymized form, and a direct inference from the recorded movement data to the identity of the participant is not intended.
Functional sensor data will initially be recorded locally on study smartphones. The data will be encrypted and transferred to GDPR-compliant servers, with server location within Germany, before further processing and storage in the TUM NAS server system. Access to the data will be restricted to authorized members of the study team. Data will be treated confidentially at all times.
The data will be stored only for as long as necessary for study conduct and scientific analysis, but for no longer than 10 years after completion of the study. If participants provide separate consent for future use of anonymized data in subsequent studies or research questions, the storage period for those data may be renewed according to the applicable study-related consent and data-protection framework.
Anonymized data may be used for scientific publications, algorithm development and evaluation, and long-term provision in research databases, provided this is covered by the applicable consent. No transfer of personal data to other institutions or organizations, either nationally or internationally, is planned.
Participants have the right to withdraw consent at any time. From the time of withdrawal, no further data will be collected. Depending on the anonymization status and consent framework, already collected anonymized data may continue to be used if they can no longer be assigned to an individual person. The lawfulness of data processing performed before withdrawal remains unaffected.
Necessity of Human Participant Research The study question requires the investigation of real human movement. Human gait is a complex neuromusculoskeletal process that cannot be adequately reproduced using simulations, artificial datasets, or mechanical models. Smartphone sensors and AI algorithms intended for gait analysis must therefore be tested under real conditions in human participants.
Furthermore, comparison with the optical motion-capture gold standard is methodologically meaningful only when both systems record movement simultaneously in the same individual. Including healthy volunteers allows a controlled technical comparison without additional disease-related confounding factors. Human participant testing is therefore necessary to reliably evaluate the technical comparability of the investigated smartphone-based gait-analysis technology.
