FDA 510(k) Search - By Innolitics (SaMD and AI Experts)

Cognivue provides clinicians in a healthcare setting with objective measurements of cognitive function as a screening aid in the assessment of adults 55 years of age and older. This is done for the purpose of identifying a potential decline in cognitive function relative to baseline test performance of other age-normal adults, referring those adults for further testing where warranted, and monitoring changes in cognitive function over time.

Cognivue presents a series of visual stimuli and a wedge-shaped cursor. The display presents stimuli at varying signal strengths with a moving domain-specific target. The patient identifies the target location by moving the cursor using the rotatory mouse manipulandum. The patient is scored based on the timing and accuracy of the responses. Repeated trials of varying difficulty characterize the subject's performance in each of the tested functional domains.

AI/ML Overview

Here's a breakdown of the acceptance criteria and the study that proves the CogniVue device meets those criteria, based on the provided text:

Acceptance Criteria and Reported Device Performance

Acceptance Criteria Category	Acceptance Criteria (from study objective/endpoints)	Reported Device Performance (from study results)
Agreement with Reference Standard (SLUMS)	Cognivue agrees with the non-reference standard, SLUMS, in identifying unimpaired and impaired test subjects.	Agreement Analysis (SLUMS and Cognivue):

Overall relationship: Deming regression showed Cognivue = -29.98 + 4.465 SLUMS with r = 0.71 (p=27 (impairment cut-off values), Cognivue related PPA = 86%, NPA = 50%, Accuracy = 0.86.
For SLUMS ≥ 27 and Cognivue ≥ 75 (unimpaired cut-off values): PPA = 66%, NPA = 67%, Accuracy = 0.66. |

Study Details

Sample sizes used for the test set and the data provenance:
- Main Clinical Validation Test Set: 401 subjects.
- Test-Retest Reliability Test Set: 358 subjects (a subset of the 401, those available for repeat testing).
- Cut-off Determination Test Set (Training Set for Cut-offs): 92 subjects.
- Data Provenance: The subjects were enrolled at 13 older adult independent living communities and adult programs for independent living seniors who are living at home. The study was prospective in its design, involving specific recruitment and protocol execution for the study. No explicit country of origin is stated, but given the FDA De Novo filing, it's highly likely the data is from the USA.
Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
- The primary ground truth for classification was the St. Louis University Mental Status (SLUMS) Examination. While the SLUMS is a clinical tool administered by a "Tester," the text doesn't specify if this "Tester" is explicitly an "expert" (e.g., a neurologist or psychologist) or a trained administrator. Given that SLUMS scores have validated cut-off scores, it relies on the established validity of the SLUMS itself, rather than real-time expert consensus for each case. The study states "Tester #2 administration of the neuropsychological test battery," implying trained personnel administered the SLUMS.
- The study used "licensed practitioners evaluating brain function" as the general user group for the device, implying that interpretation of SLUMS scores would typically fall under their purview.
Adjudication method (e.g., 2+1, 3+1, none) for the test set:
- There was no adjudication method mentioned for establishing the ground truth (SLUMS scores). The SLUMS is a standardized psychometric test with established scoring rules and cut-offs.
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 such MRMC comparative effectiveness study was conducted. The study design was a comparative study between the Cognivue device and a battery of non-reference standard tests (including SLUMS), not a study comparing human readers with and without AI assistance. The Cognivue is intended as an "adjunctive tool," meaning a clinician would use its output alongside other information, but the study did not quantify improvement of human performance.
If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:
- Yes, a standalone performance assessment was done. The entire clinical validation focuses on the Cognivue device's output (an average score classified as impaired, intermediate, or unimpaired) and its agreement with the SLUMS examination and other neuropsychological tests. The device itself generates the score and classification automatically without real-time human interpretation or modification of its core output.
The type of ground truth used (expert consensus, pathology, outcomes data, etc):
- The primary ground truth used was a validated clinical psychometric scale: the St. Louis University Mental Status (SLUMS) Examination. This is a well-established cognitive screening tool with predefined cut-off scores for unimpaired, intermediate (mildly impaired), and impaired cognitive function. Secondary ground truth involved other established neuropsychological tests for construct validity.
The sample size for the training set:
- The cut-off values for Cognivue (impaired, intermediate, unimpaired) were determined using a preliminary study of 92 subjects. This set of 92 subjects essentially served as a training set for calibrating the device's classification thresholds against the SLUMS.
How the ground truth for the training set was established:
- For the 92 subjects used to establish the Cognivue cut-offs, the ground truth was established by administering the St. Louis University Mental Status (SLUMS) Examination. The SLUMS scores were then used to classify these 92 subjects into impaired, intermediate, and unimpaired categories based on their established cut-offs (e.g., >26 unimpaired,

Summary

Regulation Number and Section

§ 882.1470 Computerized cognitive assessment aid.

(a)
Identification. The computerized cognitive assessment aid is a prescription device that uses an individual's score(s) on a battery of cognitive tasks to provide an interpretation of the current level of cognitive function. The computerized cognitive assessment aid is used only as an assessment aid to determine level of cognitive functioning for which there exists other valid methods of cognitive assessment and does not identify the presence or absence of clinical diagnoses. The computerized cognitive assessment aid is not intended as a stand-alone or adjunctive diagnostic device.(b)
Classification. Class II (special controls). Except when the computerized cognitive assessment aid is intended for diagnostic assessment of specific diseases or conditions and relies on inputs from visual cues, auditory cues, and/or functional use of the hand, the device is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 882.9. The special control(s) for this device are:(1) The technical parameters of the device's hardware and software must be fully characterized and be accompanied by appropriate non-clinical testing:
(i) Hardware specifications must be provided. Appropriate verification, validation, and hazard analysis must be performed.
(ii) Software, including any proprietary algorithm(s) used by the device to arrive at its interpretation of the patient's cognitive function, must be described in detail in the software requirements specification (SRS) and software design specification (SDS). Appropriate software verification, validation, and hazard analysis must be performed.
(2) The device must be designed and tested for electrical safety.
(3) The labeling must include:
(i) A summary of any testing conducted to demonstrate how the device functions as an interpretation of the current level of cognitive function. The summary of testing must include the following, if available: Any expected or observed adverse events and complications; any performance measurements including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) per the devices intended use; a description of the repeatability of measurements; a description of how the cut-off values for categorization of measurements were determined; and a description of the construct validity of the device.
(ii) A warning that the device does not identify the presence or absence of clinical diagnoses.
(iii) A warning that the device is not a stand-alone diagnostic.
(iv) The intended use population and the intended use environment.
(v) Any instructions technicians must convey to patients regarding the administration of the test and collection of cognitive test data.