Technology-Driven Future

Using the actuation-perception-control analytical framework, this paper systematically reviews the core technological advancements of intelligent prosthetic hip and knee joints, which closely aligns with the 'sensing-decision-making-execution' technical pipeline. The perception layer involves fusion processing of physiological signals (e.g., sEMG) and mechanical signals, while the control layer covers three key strategies: torque compensation, motion tracking, and direct intention control. The paper explicitly identifies that the primary current bottlenecks are the lack of effective real-time intention recognition methods and insufficient adaptability in dynamic environments.

Published in November 2024 in IEEE Transactions on Medical Robotics and Bionics, this systematic review comprehensively synthesizes the research progress of semi-active and active lower limb prostheses, with a core focus on three key dimensions: the mechatronic characteristics of the devices, perception and control strategies, and performance validation with end-users. Its multi-dimensional analytical framework of 'prototype technical metrics—clinical indications—target populations—market translation' is particularly valuable for demonstrating to investors the mature translation pathway of intelligent prostheses 'from lab to product'.

This systematic review, conducted in accordance with PRISMA guidelines, involved a comprehensive analysis of 60 full-text scientific articles. The review focuses on four critical areas: how intent detection is defined across various studies, the sensor technologies used and sensor placement, the machine learning methods developed for intent detection, and the metrics used to evaluate the performance of these methods. Reliable intent detection is crucial for improving adaptability, reducing the cognitive effort required from users, and maintaining safety in changing environments.

This systematic review established the world's first implantable brain-computer interface (iBCI) trial participant registry, systematically identifying 112 studies across the US, Europe, China, and Australia, recording 80 implantees in total. Nearly half (49.1%) of the publications appeared after 2020, indicating that implantable BCI is in a period of rapid development. Devices have been used to control robotic prostheses, digital devices, and other external effectors, offering authoritative data support for future BCI-integrated intelligent prostheses or exoskeleton product line extensions.

This study systematically retrieved 4 databases from 2011–2023, including 27 studies (591 spinal cord injury patients, 10 exoskeletons) from 555 articles. Exoskeletons were grouped into actively controlled (only HAL, bioelectrical signal-detecting) and passively controlled categories. Results showed: HAL significantly improved unassisted 6-minute walk distance and 10-meter walk speed, and was the only device with consistent clinically meaningful improvements across all outcomes. BCI-integrated training outperformed standalone training, validating the clinical value of actively controlled exoskeletons.