IdentiARAT: Toward Automated Identification of Individual ARAT Items from Wearable Sensors

2025-04-17

AI Summary

To address the subjectivity and time-consuming nature of clinical assessments using the Action Research Arm Test (ARAT), this paper proposes an automated, fine-grained motion recognition method based on wrist-worn inertial measurement unit (IMU) sensors. We introduce MiniROCKET—a fast, interpretable time-series classifier—for the first time in ARAT task recognition and systematically evaluate multiple signal preprocessing strategies to enhance discriminability among highly similar upper-limb actions (e.g., grasping vs. pinching). Evaluated on data from 45 participants, our approach achieves high domain-level classification accuracy (mean F1-score > 0.92) with inference speed accelerated by two orders of magnitude. The key contributions are: (1) the first application of MiniROCKET to rehabilitation motion recognition, and (2) empirical optimization of preprocessing pipelines for ARAT-specific gesture discrimination. This work establishes a novel, wearable-based paradigm for standardized, objective, and reproducible ARAT assessment.

Technology Category

• Intelligent Robots: Manipulation
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Computer Vision: Motion & Tracking

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This study explores the potential of using wrist-worn inertial sensors to automate the labeling of ARAT (Action Research Arm Test) items. While the ARAT is commonly used to assess upper limb motor function, its limitations include subjectivity and time consumption of clinical staff. By using IMU (Inertial Measurement Unit) sensors and MiniROCKET as a time series classification technique, this investigation aims to classify ARAT items based on sensor recordings. We test common preprocessing strategies to efficiently leverage included information in the data. Afterward, we use the best preprocessing to improve the classification. The dataset includes recordings of 45 participants performing various ARAT items. Results show that MiniROCKET offers a fast and reliable approach for classifying ARAT domains, although challenges remain in distinguishing between individual resembling items. Future work may involve improving classification through more advanced machine-learning models and data enhancements.

Problem

• Automate ARAT item labeling using wrist-worn sensors
• Reduce subjectivity and time in ARAT clinical assessments
• Classify ARAT items via IMU data and MiniROCKET

Innovation

• Wrist-worn IMU sensors for ARAT labeling
• MiniROCKET for time series classification
• Preprocessing strategies to enhance data information