Is wearable data reliable for monitoring behavior? Design of a wearable-based IoMT puzzle game for remote behavior monitoring
Sadhu, Shehjar; Nishtha, Bhagat; Castillo, Elijah; weyandt, lisa; Mankodiya, Kunal; Solanki, Dhaval
- Publisher
- Zenodo
- Language
- en
Abstract
Abstract— Wearable Internet of Medical Things (IoMT) platforms are transforming remote health monitoring by enabling continuous, real-world data acquisition for chronic and neurodevelopmental conditions. However, ensuring data quality and system reliability remains a key challenge, particularly outside controlled clinical environments. Attention-Deficit/Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder that affects approximately 11.8 million children in the US. While existing treatment plans can help manage symptoms, current assessment methods rely heavily on subjective clinical rating scales, which are prone to errors and may lead to inaccurate evaluations. To address this, we present MindGame, a wearable IoMT system that integrates a commercially available smartwatch (Samsung Galaxy Watch 4) and a digital puzzle game to remotely monitor ADHD related behaviors. MindGame synchronizes gameplay data with physiological signals (accelerometer, gyroscope, heart rate, and computer mouse movement. We conducted a week-long study with 5 ADHD and 7 control participants, who used the system in both remote and laboratory environments. Twenty-seven unique puzzles were created. We analyzed four key data quality metrics: signal-to-noise ratio, percentage data loss, percentage of zeros (e.g., 0 bpm heart rate), and sample rate consistency. MindGame successfully collected data from 2427 puzzles with multimodal sensors from different devices. Our results highlight critical differences in data quality across remote and laboratory settings, validating the importance of adaptive quality check algorithms for real-world sensing.
Full text
MindGame Study Dataset README In this file, we present an overview of the MindGame dataset. Figure 1 shows the structure of the MindGame file system. The following are details regarding each parameter in the file system: Figure 1: Overview of the dataset file structure collected from the MindGame system. The dataset consists of two primary data streams: Physiology and Computer Interaction. info.csv: Column Description screen_width_list, screen_height_list Screen resolution during gameplay main_level_list, sub_level_list Puzzle level and sublevel time_to_complete_level_list, total_seconds Puzzle completion time json_file_list Raw puzzle log filename sessionID_list, int_session Unique session identifiers pid, pid_numeric, group Participant ID and group (e.g., Neurotypical, ADHD) session_time AM/PM indicator number_puzzles, puzzle_number Session structure metadata computer_mouse.csv
Column Description x, y Mouse cursor coordinates (pixels) timestamp Timestamp (HH:MM:SS:MS) x(px/s^2), y(px/s^2) Approximate acceleration from mouse movement filename Mouse data file pid, group, level, sub_level, session_time Metadata for merging int_session, puzzle_number, pid_numeric Participant/session indexing hr.csv Column Description filename Raw heart rate file HR % Missing, HR % 0s Missing and zero-value percentages HR SNR Signal-to-noise ratio HR SRC Signal reliability coefficient pid, group, level, sub_level, puzzle_number Metadata for alignment Session_ID, session_time, int_session, number_puzzles, pid_numeric Session-level info acc.csv Column Description x(m/s^2), y(m/s^2), z(m/s^2) Accelerometer readings from the wrist internal_ts, watch_timestamp, relative_timestamp High-resolution timestamps filename Accelerometer data file
pid, group, level, sub_level, session_time, puzzle_number_x Metadata for alignment ts_only, ts_seconds, int_session, number_puzzles, pid_numeric Time-aligned fields for multimodal fusion Notes: This dataset is intended solely for research and educational purposes. All data collection was conducted under IRB-approved protocols with informed consent.
