Construction of an AI System for the Automatic Supervision of Shoulder's Rehabilitation Exercises (Rehab-SPIA) (Rehab-SPIA)

Construction of an Artificial Intelligence System for the Remote Automatic Supervision of Shoulder's Rehabilitation Exercises

The current historical phase and the growing need for rehabilitation in the world make tele-rehabilitation systems, and e-Health in general, fundamental tools for increasing patient engagement and compliance with care, crucial elements for the preservation of the NHS from a perspective expenditure review and resource optimization. In particular, the rehabilitation patient has on average an adherence to the Home Exercise Program (HEP) between 30-50%, to which is frequently added a reduced effectiveness of motor learning due to the lack of feedback on the accuracy of the gesture, as is the case. it happens in the hospital or outpatient setting under the supervision of a therapist.

The new computational approaches for the analysis of data on human movement, aimed at the development of algorithms to automatically supervise the accuracy of the patient's gesture during home self-treatment exercise such as those based on Artificial Intelligence (AI) and Machine Learning (ML), especially those of the latest generation, called sub-symbolics (or connectionists) can help.

Among the most promising approaches are. Given the importance of the Home Exercise Program in shoulder disease, it was decided to select a population of patients affected by the main pathologies affecting this joint.

The main objective of the study is to create and validate a software tool for the automatic and expert analysis of the correct execution of the main rehabilitation exercises for the functional recovery of the shoulder following orthopedic pathologies.

Study Overview

• Status: Recruiting
• Conditions: Rotator Cuff Tears
• Intervention / Treatment: Diagnostic test: Pathologic Exercise

Detailed Description

The current historical phase and the growing need for rehabilitation in the world make tele-rehabilitation systems, and e-Health in general, fundamental tools for increasing patient engagement and compliance with care, crucial elements for the preservation of the NHS from a perspective expenditure review and resource optimization .

In particular, the rehabilitation patient has on average an adherence to the Home Exercise Program (HEP) between 30-50%, to which is frequently added a reduced effectiveness of motor learning due to the lack of feedback on the accuracy of the gesture, as it happens in the hospital or outpatient setting under the supervision of a therapist.

The new computational approaches for the analysis of data on human movement, aimed at the development of algorithms to automatically supervise the accuracy of the patient's gesture during the exercise of home self-treatment, attempt to solve this last critical issue.

Among the most promising approaches are those based on Artificial Intelligence (AI) and Machine Learning (ML), in particular those of the latest generation, called sub-symbolic (or connectionist).

These algorithms arouse a lot of interest for their ability to automatically extract the salient properties of the movement, reducing the intervention of experts to the collection of all the data, and to the possible labeling of the examples (5) In any case, the literature shows a lack of models developed with the direct involvement of clinicians and a scarcity of data sets created with patient populations.

Furthermore, most of the models present in the literature have been created using numerous input devices, often with a high technological rate with considerable costs for implementing a possible service at the patient's home.

For these reasons we want to create a specialist clinical dataset, starting only from the videos of the exercises, involving specific populations by pathology and built on the basis of clinical judgment. With these characteristics, this project aims to automate the motion analysis process as much as possible, enormously reducing the costs deriving from the use of technologies and minimizing human error, all by exploiting the most recent computational approaches in order to create a useful and low-cost tool for home functional re-education.

Given the importance of the Home Exercise Program in shoulder disease, it was decided to select a population of patients affected by the main pathologies affecting this joint.

The main objective of the study is to create and validate a software tool for the automatic and expert analysis of the correct execution of the main rehabilitation exercises for the functional recovery of the shoulder following orthopedic pathologies.

• Study Type: Observational
• Enrollment (Estimated): 100

Contacts and Locations

• Study Contact: Name: Maria Grazia Benedetti, MD; Phone Number: +390516366236; Email: benedetti@ior.it

Study Locations

• Italy
  • Bologna, Italy, 40136
    • Recruiting
    • IRCCS-Istituto Ortopedico Rizzoli
    • Contact: Maria Grazia Benedetti, MD; Phone Number: +39051 6366236; Email: benedetti@ior.it

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years to 65 years (Adult, Older Adult)
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

Patients after arthroscopic reconstruction of rotator cuff

Description

Inclusion Criteria:

• Healthy subjects group:

  • Adult patients> 18 years old
  • Patients with no known shoulder pathologies

• Group of subjects with shoulder pathology operated on

  • Adult patients> 18 years
  • Suffering from orthopedic pathologies affecting the shoulder such as: outcomes of ultrasound-guided percutaneous treatment for tendon calcification, outcomes of ultrasound-guided detachment in adhesive bursitis, outcomes of proximal humerus fractures, repair of the rotator cuff, interventions for scapulo-humeral instability.

Exclusion Criteria:

• Patients with a history of opioid drug dependence or a history of substance abuse
• Patients suffering from orthopedic pathologies affecting the upper limbs in the presence of clear detectable surgical complications
• Patients with cognitive disorders (MMSE Mini Mental State Examinantion greater than or equal to 24/30).
• Patients suffering from major anamnestic or current neurological or psychiatric pathologies, severe cardiopulmonary, hepatic or renal pathologies that contraindicate participation in the study.

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Healthy; Healthy subjects, wothout shoulder pathology	Diagnostic test: Pathologic Exercise; In the first phase of the project, a series of 5 active shoulder mobilization exercises characterized by an adequate range of motion will be tested to verify the set up for the video recording.; In the second phase shoulder movements will be recorder by a smartphone.; A questionnaire will be used and adapted on the basis of which to evaluate the correctness of the exercises performed by each healthy subject / patient. This questionnaire will provide a Clinical Score (CS) which assigns a numerical value to the patient's overall performance for each repetition.; The videos of each repetition of exercises performed by the healthy subjects / patients will then be evaluated by two different clinicians, blinded, using the questionnaire.; The Artificial Intelligence learning algorithm will be able to output an evaluation score that will be compared with that produced by clinicians.
Rotator cuff tears; Patients after arthroscopic reconstruction of rotator cuff	Diagnostic test: Pathologic Exercise; In the first phase of the project, a series of 5 active shoulder mobilization exercises characterized by an adequate range of motion will be tested to verify the set up for the video recording.; In the second phase shoulder movements will be recorder by a smartphone.; A questionnaire will be used and adapted on the basis of which to evaluate the correctness of the exercises performed by each healthy subject / patient. This questionnaire will provide a Clinical Score (CS) which assigns a numerical value to the patient's overall performance for each repetition.; The videos of each repetition of exercises performed by the healthy subjects / patients will then be evaluated by two different clinicians, blinded, using the questionnaire.; The Artificial Intelligence learning algorithm will be able to output an evaluation score that will be compared with that produced by clinicians.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Correctness of the shoulder movement	A questionnaire in which the clinician will describe the correctenss of the shoulder movement will be used and compared with the attribution by the Artificial Intelligence software	12 months

Collaborators and Investigators

• Sponsor: Istituto Ortopedico Rizzoli
• Investigators: Principal Investigator: Maria Grazia Benedetti, MD, Istituto Ortopedico Rizzoli

Study record dates

Study Major Dates

• Study Start (Actual): April 1, 2020
• Primary Completion (Estimated): September 30, 2024
• Study Completion (Estimated): September 30, 2024

Study Registration Dates

• First Submitted: August 27, 2021
• First Submitted That Met QC Criteria: August 27, 2021
• First Posted (Actual): August 30, 2021

Study Record Updates

• Last Update Posted (Actual): March 27, 2024
• Last Update Submitted That Met QC Criteria: March 25, 2024
• Last Verified: March 1, 2024

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Wounds and Injuries; Rupture; Shoulder Injuries; Tendon Injuries; Rotator Cuff Injuries
• Other Study ID Numbers: 0002017

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No