Search Results

NeuroRPM is intended to quantify movement disorder symptoms during wake periods in adult patients 46 to 85 years of age with Parkinson's disease. These symptoms include tremor, bradykinesia, and dyskinesia. NeuroRPM is intended for clinic and home environments.

NeuroRPM is a software application for the Apple Watch that is prescribed by a health professional to quantify motor symptoms of Parkinson's disease including bradykinesia, dyskinesia, and tremor. NeuroRPM collects accelerometer and gyroscope data from the Apple Watch. The motion data are transmitted to cloud servers and analyzed using machine learning models developed to generate binary symptom classifications. Binary symptom classification output is generated every 15-minutes.

Here's a breakdown of the acceptance criteria and study details for the NeuroRPM device, based on the provided text:

1. Table of Acceptance Criteria & Reported Device Performance

The acceptance criteria for NeuroRPM were based on achieving specific sensitivity and specificity thresholds for detecting tremor, bradykinesia, and dyskinesia. While explicit "acceptance criteria" values (e.g., "must meet X sensitivity") aren't directly stated as minimum required values, the study design aimed to demonstrate adequate performance to support its intended use and substantial equivalence to the predicate device. The reported performance is presented with 95% confidence intervals.

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

• Sample Size for Test Set: The study was conducted with 36 subjects. However, the analysis was based on "events" rather than subjects.
  • Total number of events (from ground truth) for sensitivity evaluation:
    • Tremor: 170
    • Bradykinesia: 203
    • Dyskinesia: 73
  • Total number of events (from ground truth) for specificity evaluation:
    • Tremor: 325
    • Bradykinesia: 292
    • Dyskinesia: 422
• Data Provenance: The study was an "observational, non-intervention study." The subject demographics indicate the study took place at a single site and had 95.5% Caucasian subjects. This suggests the data is retrospective in terms of being collected from past observations, but the study itself was designed prospectively to collect this data for validation. The country of origin is not explicitly stated beyond being a "single site," but given the FDA submission, it implicitly aligns with U.S. regulatory standards.

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

• Number of Experts: 3
• Qualifications of Experts: Board-certified movement disorder specialists.

4. Adjudication Method for the Test Set

• Adjudication Method: The ground truth for each sample was derived based on the majority score of the expert rater panel. This implies a 2-out-of-3 or 3-out-of-3 consensus approach.

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

• No, an MRMC comparative effectiveness study was not done. The study evaluated the standalone performance of the NeuroRPM device in quantifying symptoms against expert-derived ground truth. It did not assess human reader performance with or without AI assistance.

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

• Yes, a standalone performance study was done. The described clinical performance testing evaluates "NeuroRPM's ability to quantify Parkinson's symptom presence or absence" directly, without human intervention in the device's output interpretation or decision-making.

7. The Type of Ground Truth Used

• Type of Ground Truth: Expert Consensus based on clinical scales. The experts provided scores using the Unified Parkinson Disease Rating Scale (UPDRS) and the Abnormal Involuntary Movement Scale (AIMS). The device's binary classifications (e.g., "Tremor detected" vs. "No tremor detected") were then mapped to specific score ranges on these validated clinical scales, and the majority score from the expert panel served as the ground truth.

8. The Sample Size for the Training Set

• The document does not explicitly state the sample size for the training set. It mentions a "machine learning model developed to generate binary symptom classifications" but does not detail the dataset used for training. The clinical performance data provided (n=36 subjects) is specifically for the validation or test set.

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

• The document does not explicitly state how the ground truth for the training set was established. While it mentions machine learning models were used, it does not provide details on the data used for training these models or how their ground truth was determined.

Ask a specific question about this device