Search Results

The VisualEyes system provides information to assist in the nystagmographic evaluation, diagnosis and documentation of vestibular disorders. Nystagmus of the eye is recorded by use of a goggle mounted with cameras. These images are measured, recorded, displayed and stored in the software. This information then can be used by a trained medical professional to assist in diagnosing vestibular disorders. The target population for VisualEyes system is 5 years of age+

VisualEyes 505/515/ 525 is a software program that analyzes eye movements recorded from a camera mounted to a video goggle. A standard Video Nystagmography (VNG) protocol is used for the testing. VisualEyes 505/515/ 525 is and update/change, replacing the existing VisualEyes 515/525 release 1 (510(k) cleared under K152112). The software is intended to run on a Microsoft Windows PC platform. The "525" system is a full featured system (all vestibular tests as listed below) while the "515" system has a subset of the "525″ features. "505″ is a simple video recording mode. The VisualEyes 505/ 515/ 525 software is designed to perform the following vestibular tests: Spontaneous Nystagmus Test, Gaze Test, Smooth Pursuit Test, Saccade Test, Optokinetic Test, Dix-Hallpike, Positional Test, Caloric Test, SHA, Step, Visual VOR, VOR Suppression, Visual Eyes 505. The system consists of a head mounted goggle/mask, a camera unit and a software application running on a standard PC.

The provided text describes a 510(k) submission for the Interacoustics VisualEyes device. The focus of the performance tests is on demonstrating substantial equivalence to predicate devices rather than proving the device meets specific acceptance criteria for a novel AI/ML algorithm.

Therefore, many of the typical acceptance criteria and study details relevant to AI/ML device performance (like sensitivity, specificity, AUC, human-in-the-loop performance, expert ground truth establishment for a novel algorithm, etc.) are not explicitly mentioned in this document. The document focuses on showing that the new version of VisualEyes (revision 2) performs similarly to its predecessor (revision 1) and another cleared predicate device (VIDEO EYE TRAKKER).

However, I can extract the relevant information from the provided text as best as possible, interpreting the "acceptance criteria" in this context as the demonstration of substantial equivalence through comparable performance.

Here's the breakdown based on the provided document:

Acceptance Criteria and Reported Device Performance for Substantial Equivalence

Since this is a 510(k) submission demonstrating substantial equivalence to a predicate device, the "acceptance criteria" can be interpreted as the demonstration that the "key algorithms for detecting and analysing nystagmus" are similar between the new device and the predicate devices. The reported performance is the qualitative finding of "equivalence."

Table of Acceptance Criteria and Reported Device Performance

Study Details for Demonstrating Substantial Equivalence

1. Sample size used for the test set and the data provenance:

   • The document states: "The demonstration was carried out as a side by side comparison where the same patient was analysed by the subject device and the predicate device simultaneously."
   • It also mentions: "All tests were performed on test subjects with conjugate eye movements."
   • However, the exact number of patients/subjects in the test set is not specified.
   • Data provenance (country of origin, retrospective/prospective): Not specified.

2. Number of experts used to establish the ground truth for the test set and the qualifications of those experts:

   • The document does not describe ground truth established by experts in the typical sense for AI/ML performance validation (e.g., for disease diagnosis).
   • Instead, the "ground truth" for this substantial equivalence study appears to be the output of the predicate device, as the goal was to show that the new device's processing ("IA Curve Tracker") yielded "high correlation" with and "equivalence" to the predicate device's output. The predicate device itself is implicitly considered the "truth" for comparative purposes.

3. Adjudication method (e.g. 2+1, 3+1, none) for the test set:

   • Not applicable as the comparison was directly between the output of the subject device and the predicate device, not against a human-adjudicated ground truth.

4. 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:

   • No MRMC study was described. The study focused on the algorithmic comparison between devices, not human reader performance with or without AI assistance. The device's purpose is to assist a trained medical professional, but its performance was validated on the algorithm's output comparison.

5. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:

   • Yes, this appears to be the primary method of validation for the algorithm. The document states: "One camera recorded the left eye and was processed in the predicate device and the other recorded the right eye and was processed in subject device." This indicates a direct comparison of the algorithmic output, separate from human interpretation or human-in-the-loop performance.

6. The type of ground truth used (expert consensus, pathology, outcomes data, etc):

   • For the algorithmic comparison, the "ground truth" was effectively the output of the predicate device. The study aimed to show that the subject device's processing produced equivalent results to the cleared predicate device for the "key algorithms for detecting and analysing nystagmus."
   • For the overall clinical performance, it states "These activities, testing and validation show that VisualEyes 505/515/ 525 perform as specified and is safe and effective," which implies a broader clinical assessment but no specific details on how "truth" was established for clinical outcomes beyond direct comparison to the predicate.

7. The sample size for the training set:

   • Not applicable/Not mentioned. This document describes a 510(k) submission for an update to an existing device, focusing on substantial equivalence tests, not the development or training of a de novo AI/ML model. The algorithm is stated as being "the same" as the predicate's "IA Curve tracker."

8. How the ground truth for the training set was established:

   • Not applicable/Not mentioned for the reasons stated above.

Ask a specific question about this device