MANA: Designing and Validating a User-Centered Mobility Analysis System
Boyd Anderson, Shenggao Zhu, Ke Yang, Jian Wang, Hugh Anderson, Chao Xu Tay, Vincent Y. F. Tan, Ye Wang · 2018 · Proceedings of the 20th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2018) · doi:10.1145/3234695.3236340
Summary
This paper presents MANA (Mobility ANAlytics), a wearable sensor system designed to measure gait parameters in people with Parkinson's Disease (PD) in clinical and home settings. The system addresses a significant gap in accessible health monitoring: existing gait analysis technologies like GAITRite pressure mats and Vicon motion capture systems cost tens of thousands of dollars, require dedicated laboratory space, and demand technical expertise — making them inaccessible to most patients and community hospitals. MANA consists of three components: custom-built IMU (Inertial Measurement Unit) sensors embedded in shoes and worn at the waist, a mobile phone app (Collator) that collects sensor data via Bluetooth Low Energy, and a web-based Hub for data analysis and visualization. A central design principle was reducing the stigma of assistive technology use. After consultations with clinicians and people with PD revealed that ankle-mounted sensors caused embarrassment, the team relocated sensors inside shoes under the arch of the foot, making them completely invisible during use. Each sensor costs less than USD , compared to tens of thousands for clinical alternatives. The system went through six hardware iterations to achieve a compact, durable design with battery life lasting up to two weeks in daily use. The authors developed novel gait analysis algorithms that fuse accelerometer and gyroscope data, leveraging kinematic constraints enabled by the unique shoe-embedded sensor placement.
Key findings
MANA was validated through a clinical trial at Huashan Hospital in Shanghai involving 60 participants: 40 people with PD across four disease stages, 10 people with rapid eye movement sleep behaviour disorder (a prodromal condition for PD), and 10 healthy age-matched controls. Participants walked an 8-meter track 8-10 times while wearing three sensors. The machine learning algorithm (SL-ML) achieved a mean absolute error of 4.0 cm for stride length and 2.6 cm for step length, outperforming all previously published IMU-based methods that included PD data. The dataset was notably diverse, with stride lengths ranging from 29 to 159 cm — reflecting the wide range from severe parkinsonian shuffling gait to healthy walking. Performance was slightly lower for severe PD cases (4.23 cm MAE vs 3.80 cm for mild), which is expected given the more erratic gait patterns. The system successfully captures complete gait cycle data including temporal parameters like cadence, step time variability, and swing/stance time ratios, providing clinicians with rich diagnostic information beyond what single-point measurement systems can offer.
Relevance
MANA demonstrates how user-centered design can make clinical health monitoring technology genuinely accessible. The paper's most important contribution to accessibility practice is its treatment of stigma as a first-class design constraint — relocating sensors from visible ankle mounts to invisible shoe inserts based on direct feedback from users with PD. This design decision, while seemingly simple, required completely rethinking the algorithms and hardware. The system's low cost, portability, and home-use capability address real barriers that prevent people with PD from accessing regular gait monitoring, particularly those in rural areas or developing countries far from specialized hospitals. For accessibility practitioners, MANA illustrates how wearable assistive technology must balance technical accuracy with social acceptability and practical usability to achieve real-world adoption.
Tags: Parkinson's disease · wearable sensors · gait analysis · inertial measurement units · mobility assessment · user-centered design · machine learning · assistive technology