2025 ISAKOS Congress in Munich, Germany

2025 ISAKOS Biennial Congress Paper


A Pilot Study for Assessing NAO Humanoid Robot Assistance in Shoulder Rehabilitation

Alessandra Raso, M.Sc., Rome ITALY
Martina Pulcinelli, M.Sc., Rome ITALY
Emiliano Schena, Eng, Rome, --- Select One --- ITALY
Alfio Puglisi, PhD, Rome ITALY
Giovanni Pioggia, PhD, Messina ITALY
Arianna Carnevale, PhD, Roma, Roma ITALY
Umile Giuseppe Longo, MD, MSc, PhD, Prof., Rome ITALY

University Campus BioMedico of Rome, Rome, ITALY

FDA Status Not Applicable

Summary

This study assesses the nao humanoid robot's performance as a guide in shoulder rehabilitation — evaluating the Range of Motion (ROM), timing, and smoothness of exercises. The robot enhances movement's temporal control, crucial for movement quality during rehabilitation

Abstract

Background

The shoulder joint, the most mobile and unstable joint in the human body, is highly susceptible to injuries like rotator cuff tears. Rehabilitation, progressing from passive to active and strengthening exercises, is crucial for restoring function and mobility post-surgery. Robotics, particularly socially assistive robots like NAO (Aldebaran Robotics), have become important in rehabilitation due to their recovery benefits. As highlighted in major clinical studies, NAO, effective in educational and rehabilitative settings, has significantly boosted patient engagement.
AIMS
This study aims to explore the potential of NAO in guiding patients through rehabilitative exercises using external audiovisual stimuli. Specifically, we assess its effectiveness in restoring temporospatial control, achieving prescribed ROM and adherence to standard execution times, and the fluidity of movement—areas less explored in previous research.

Methods

A preliminary analysis involved ten healthy volunteers and two patients with shoulder musculoskeletal disorders at six months of rehabilitation follow-up. The study focused on two specific phases of the rehabilitation protocol, involving exercises such as flexion-extension (FE), external rotation (ER), and internal rotation (IR), performed at two different speeds to accommodate participant conditions. Volunteers completed these exercises for both protocol phases at both speeds, while patients completed exercises for one phase at one speed, all under therapist supervision. Kinematic analysis was performed in Visual 3D software after pre-processing markers trajectories acquired with the Qualisys™ stereophotogrammetric system. Finally, the kinematic results obtained in Visual 3D were analyzed in MATLAB®. The temporospatial control was quantified by analyzing two key parameters: the Mean Absolute Error (MAE) of the ROM achieved to evaluate adherence to the clinical protocol and the MAE of time, which measures the consistency between the assigned and actual execution times for each exercise repetition. In all analyses, a comparison was made between the parameters recorded during a rehabilitation session with the assistance of the robot and a second session without NAO.

Results

Contrary to the results from ROM and smoothness, preliminary findings on exercise duration show a primary benefit of using NAO in rehabilitation sessions. Its ability to ensure patients adhere to predetermined time frames by mimicking its movements is demonstrated by a significant reduction in the MAE for timing with NAO versus without it. Specifically, for the FE exercises performed by the patients, the MAE decreased from 5.0 s without NAO to 0.4 s with NAO. For ER exercises performed by the patients, MAE reduced from 1.0 s without NAO to 0.3 s with NAO. For IR exercises performed by the patients, MAE reduced from 1.0 s without NAO to 0.2 s with NAO.

Discussion

As previously mentioned, the factors we investigated—including ROM, timing, and smoothness—have been less studied in other research, leading to our main finding, i.e., an enhancement in temporal control. This aspect is of significant clinical importance as it directly affects the quality of movement. These encouraging initial outcomes suggest the need for additional research to expand and validate these findings.

This work was supported by the Italian Ministry of Research, under the complementary actions to the NRRP “Fit4MedRob - Fit for Medical Robotics” Grant (# PNC0000007) - CUP: B53C22006980001