Validation of a Mobile, Sensor-based Neurobehavioral Assessment With Digital Signal Processing and Machine-learning...

Area of Science:

• Digital health
• Neuroscience
• Gerontology

Background:

• The Miro Health Mobile Assessment Platform utilizes self-administered assessments for neurobehavioral and cognitive evaluation.
• These assessments measure behaviors typically assessed by specialized clinicians.

Purpose of the Study:

• To assess the concurrent validity and test-retest reliability of the Miro Health Mobile Assessment Platform.
• To evaluate the platform's performance in classifying mild cognitive impairment (MCI).

Main Methods:

• Concurrent validity was assessed using Spearman correlations between Miro Health variables and legacy neuropsychological test scores.
• Test-retest reliability was quantified using intraclass correlations.
• A machine-learning algorithm generated a Risk score to differentiate healthy controls (HC) from MCI and its subtypes (amnestic MCI [aMCI] and nonamnestic MCI [naMCI]).

Main Results:

• Miro Health variables showed moderate to strong correlations (0.27-0.68) with legacy test scores in HC.
• Test-retest reliabilities ranged from 0.25-0.79, with minimal learning effects observed.
• The Risk score demonstrated high accuracy in differentiating aMCI from HC (AUROC=0.97), naMCI from HC (AUROC=0.80), combined MCI from HC (AUROC=0.89), and aMCI from naMCI (AUROC=0.83).

Conclusions:

• The Miro Health Mobile Assessment Platform provides a valid and reliable method for assessing neurobehavioral and cognitive status.
• The platform effectively distinguishes between healthy controls and individuals with MCI.
• The platform can differentiate between subtypes of MCI, specifically amnestic and nonamnestic MCI.