Search Results

The GEMS-H is a robotic exoskeleton that fits orthotically on the wearer's waist and thighs, outside of clothing. The device is intended to help assist ambulatory function in rehabilitations under the supervision of a trained healthcare professional for the following population:

· Individuals with stroke who have gait deficits and exhibit gait speeds of at least 0.4 m/s and are able to walk at least 10 meters with assistance from a maximum of one person.

The trained healthcare professional must successfully complete a training program prior to use of the device. The device is not intended for sports.

The GEMS-H is a lightweight, robotic exoskeleton designed to help assist ambulatory function of stroke patients who meet the assessment criteria, in rehabilitations under the supervision of a trained healthcare professional. The GEMS-H device provides assistance to the patient during hip flexion and extension.

The device is worn over clothing around the wearer's waist and fastened with Velcro straps to assists hip flexion and extension. The device weighs 4.7 lbs (2.1 kg) and has two motors that run on a single rechargeable battery. The device is equipped with joint angle and electrical current sensors to monitor hip joint angle and torque output, respectively.

The assist torque is transmitted to the wearer's thighs via thigh support frames. A trained healthcare professional, who operates the device, can change assist settings through software that runs on the tablet PC.

This document describes the premarket notification (510(k)) for the Samsung GEMS-H, a powered lower extremity exoskeleton. The information provided primarily focuses on establishing substantial equivalence to a predicate device, rather than proving the device meets specific acceptance criteria related to an AI's performance.

Based on the provided text, the device itself (GEMS-H exoskeleton) is the subject of the regulatory review, and the "study" described is a clinical trial to assess its safety and effectiveness in assisting ambulatory function in stroke patients. There is no mention of an AI component requiring specific performance acceptance criteria for an algorithm or model.

Therefore, many of the requested points regarding AI acceptance criteria, ground truth establishment, expert adjudication, and MRMC studies are not applicable directly to this document's content, as it's not about an AI-powered diagnostic or predictive device.

However, I can extract information related to the device's clinical performance and the study design:

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria for the device are defined in terms of safety and effectiveness, based on a clinical trial.

Note: The primary effectiveness endpoint reported (+0.12 m/s) did not explicitly meet the stated objective of "≥0.14 m/s". However, the FDA's clearance indicates that the overall evidence was sufficient for substantial equivalence. The document highlights that the exploratory endpoints showed larger improvements with the device, and that the "study subjects achieved a mean clinically significant improvement (MCID) in gait speed as measured by the 10MWT, and in walking endurance as measured by the 6MWT."

2. Sample Size Used for the Test Set and Data Provenance

• Test Set Sample Size: The clinical study enrolled 53 subjects, of whom 41 completed the entire protocol. This effectively serves as the "test set" for the device's performance.
• Data Provenance: The study was conducted in the United States (Shirley Ryan AbilityLab, Northwestern University, Chicago, IL). It was a prospective, single arm, interventional, open-label, single center study.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts

This question is not applicable in the context of this device. The "ground truth" for the device's performance is objective physiological measurements (gait speed, walking endurance) and adverse event reporting, observed by medical professionals (licensed physical therapists and physicians) during rehabilitation sessions. It's not a diagnostic AI where experts label images or data for truth.

4. Adjudication Method for the Test Set

Not applicable. This is not an imaging or diagnostic study requiring adjudication of expert interpretations.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

Not applicable. This device is an exoskeleton for physical assistance, not an AI for human reader enhancement.

6. If a Standalone (i.e. algorithm only without human-in-the-loop performance) was done

This question is not directly applicable. The device itself performs its function (providing mechanical assistance), and its performance is evaluated with human-in-the-loop (the patient wearing it and being supervised by a healthcare professional). The document states the device is intended to assist ambulatory function "under the supervision of a trained healthcare professional." There is no "standalone algorithm" performance to report in this context.

7. The Type of Ground Truth Used

The "ground truth" for this device's performance was based on:

• Objective functional mobility measures: 10-Meter Walk Test (10MWT) for gait speed and 6-Minute Walk Test (6MWT) for walking endurance. These are standardized, objective measures of physical performance.
• Adverse Event reporting: Clinical observation and documentation of any adverse events during the study.

8. The Sample Size for the Training Set

Not applicable, as this is not an AI model requiring a training set in the typical sense. The clinical study of 41 subjects who completed the protocol served as the primary evidence.

9. How the Ground Truth for the Training Set was Established

Not applicable for the same reason as #8. The data collected from the 41 patients in the clinical study are the direct evidence of the device's performance in a real-world (rehabilitation) setting.

Ask a specific question about this device