Search Results

The BrainPulse is intended for use on a patient's head to non-invasively detect, amplify and capture the skull motion caused by pulsatile flow from the cardiac cycle. The BrainPulse is not in the diagnosis of neurological conditions, diseases, or disorders.

The BrainPulse™ 1100 is a non-invasive device having three basic components that are provided as an entire system and all are non-sterile and reusable: . Headset with detachable cable that is patient contacting and includes: One Photoplethysmograph (PPG) sensor for o detecting the heart rate and timing; One Sound Pressure Level (SPL) sensor for o detecting ambient environment noise; and Six Accelerometer sensors to detect the o acceleration at six selected locations. ● Data Collector, which digitizes the analog signals from the headset . Computer, which incorporates the device Software and space to store the BrainPulse recording data and, . Device Software, which provides the user interface, hardware control software libraries The BrainPulse collects and stores skull motion caused by pulsatile flow from the cardiac cycle. The normal brain structure produces a motion that is driven by and synchronized with the heart rate, and is manifested by slight acceleration of the skull. The device uses piezoelectric-based accelerometer sensors that measure skull motion rather than brain sounds. The typical frequencies that are employed with the BrainPulse are below 20 Hz and mostly below 10 Hz, well below the lower limit of audible sound. The headset senses the motion and the Data Collector digitizes the signal. The computer mainly provides the user interface and stores the data for further processing. Users place the device's headset on patients and setup the user interface to perform a BrainPulse recording. Typically, a recording is about 2-3 minutes long, though users may obtain recordings up to 30 minutes long depending on their applications. Recordings may be obtained at any time.

The provided text describes the BrainPulse 1100 device and modifications to its software, but it does not contain details about specific acceptance criteria or a study proving the device meets those criteria, especially in terms of diagnostic performance or clinical effectiveness.

Here's a breakdown of what can be extracted and what information is missing:

1. Table of Acceptance Criteria and Reported Device Performance:

• Missing from the provided text. The text mentions "The device met all acceptance criteria during verification and validation tests and demonstrated compliance to design inputs," but it does not specify what those acceptance criteria were or present quantitative performance data.

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

• Missing from the provided text. The text mentions "Software validation as well as system verification was performed," but does not detail the sample size or the nature of the test set (e.g., number of recordings, patients, data provenance).

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

• Missing from the provided text. There is no mention of expert involvement in establishing ground truth for any testing.

4. Adjudication method for the test set:

• Missing from the provided text. No adjudication method is described.

5. If a multi-reader, multi-case (MRMC) comparative effectiveness study was done:

• Missing from the provided text. There is no mention of an MRMC study or any comparison of human readers with or without AI assistance. The device's intended use is to "detect, amplify and capture the skull motion," not to aid human interpretation of images or other data related to neurological conditions.

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

• Partially addressed, though not in terms of clinical performance. The text indicates that "Software validation as well as system verification was performed to show equivalence to the previous software that was cleared in DEN140040." This suggests an evaluation of the algorithm's output in isolation to ensure it matches the previous version, but not in the context of clinical accuracy or standalone diagnostic performance. The device itself is described as collecting and storing data for "further processing," implying it's a data acquisition tool, not a diagnostic algorithm.

7. The type of ground truth used:

• Missing from the provided text for performance evaluation. For the software validation, the implicitly used "ground truth" was likely the expected output of the previous software version, as the goal was to demonstrate equivalence. However, this is not a clinical ground truth like pathology or outcomes data.

8. The sample size for the training set:

• Not applicable / Missing from the provided text. The document describes a device (BrainPulse 1100) and minor software updates, not the development of a machine learning algorithm. Therefore, there's no mention of a training set for an AI model.

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

• Not applicable / Missing from the provided text. As above, this document does not describe the development or training of an AI algorithm.

Summary of Device and Changes (from the provided text):

The BrainPulse 1100 is a non-invasive cranial motion measurement device designed to detect, amplify, and capture skull motion caused by pulsatile flow from the cardiac cycle. It consists of a headset with sensors (PPG, SPL, six accelerometers), a data collector, a computer with device software.

The current submission (K170926) is for the same device and model (BrainPulse 1100) that was previously cleared under de novo DEN140040. The only change is to the software user interface. The hardware remains unchanged.

Software Updates:

• Improved user interface with options to add metadata (hospital, patient details).
• Options to view the signal in real-time.
• Options to review the signal after recording.
• Automatic stopping of recording based on specified time and heartbeat intervals.
• Calculation and display of statistics from the signal for evaluation.

Testing Conducted:

• Software validation and system verification were performed to show equivalence to the previous software cleared in DEN140040.
• Risk analysis was conducted on the updated software.
• Overall system verification was performed.
• The software validation had a Level of Concern determined to be moderate (Level B).
• The device "met all acceptance criteria during verification and validation tests and demonstrated compliance to design inputs."

Crucially, the document explicitly states: "The BrainPulse is not indicated to aid in the diagnosis of neurological conditions, diseases, or disorders." This indicates that its purpose is data acquisition, not diagnostic interpretation, which explains the absence of clinical performance metrics typically associated with diagnostic AI.

Ask a specific question about this device

The BrainPulse is intended for use on a patient's head to non-invasively detect, amplify and capture the skull motion caused by pulsatile flow from the cardiac cvcle. The BrainPulse is not indicated to aid in the diagnosis of neurological conditions, diseases, or disorders.

As described above, the BrainPulse, Model 1100 (BrainPulse) is designed to measure skull motion caused by pulsatile blood flow. The BrainPulse measures these cranial pulsatile movements via an array of accelerometers placed on the scalp. The system consists of three main components: a headset, data collector, and computer.

The headset contains a forehead photoplethysmograph (PPG) sensor that measures the patient's pulse rate, a Sound Pressure Level (SPL) sensor for detecting ambient environment noise, and six accelerometers to detect the acceleration of the skull at six selected locations. These acceleration measurements typically fall in the range of 0.001 - 0.03 g.

The data collector converts the analog signals from the headset sensors and provides a digital data stream via Ethernet cable to the computer. The computer is loaded with software that allows for the user to initiate and end recordings and to manage saved data files. The BrainPulse software is not capable of displaying the recorded data from the headset; rather the data are saved in multiple file formats that can be readily displayed using other third-party software for post-hoc review.

Here's an analysis of the acceptance criteria and the study that proves the BrainPulse, Model 1100 meets these criteria, based on the provided text.

Based on the provided text, the BrainPulse, Model 1100 is a Class II device intended to non-invasively detect, amplify, and capture skull motion caused by pulsatile flow from the cardiac cycle. It is not intended for diagnostic purposes. Therefore, the "acceptance criteria" for this device are primarily focused on its ability to accurately, precisely, stably, and repeatably measure cranial motion, and its safety, rather than diagnostic performance metrics like sensitivity, specificity, or AUC.

The information provided describes the assessment of the device against a set of standards and performance expectations, rather than a single "study" with a specific test set, ground truth, and expert adjudication as might be seen for a diagnostic AI algorithm.

1. Table of Acceptance Criteria and Reported Device Performance

Given the nature of this device (a measurement tool, not a diagnostic one), the "acceptance criteria" are derived from the "Special Controls" and the objectives of the various non-clinical and clinical performance tests.

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

As this device is a measurement tool and not a diagnostic AI, there isn't a dedicated "test set" in the sense of a validation dataset for a diagnostic algorithm. Instead, its performance was assessed through various bench and clinical evaluations.

• Clinical Performance Data: 616 successful recordings from 273 patients across six clinical studies.
• Data Provenance: Studies were conducted "both at centers within and outside the United States." The text also mentions that summaries of these studies were "supplied to support a determination of a reasonable assurance of the safety and effectiveness." It is not specified whether these were specifically prospective or retrospective studies for the purpose of regulatory submission, but rather they appear to be existing clinical studies from which data was leveraged.

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

For a device like BrainPulse, which measures a physical phenomenon (skull motion due to pulsatile flow), the concept of "ground truth" for a diagnostic outcome established by human experts is not directly applicable.

• Ground Truth for Measurement Performance: The "ground truth" for evaluating the measurement capabilities of the accelerometers and PPG sensor would have been established by:

  • Reference Measurement Devices: For accelerometer resolution and SPL sensor accuracy tests, comparison to "reference measurement device" was used.
  • Known Physical Inputs/Conditions: Accelerometer stability/repeatability and hardware verification would likely involve known mechanical inputs or environmental conditions to test the device's output.
  • Physiological Correlation: For the PPG sensor, performance was evaluated by "visual comparison to concurrent SpO2 (blood oxygen saturation) sensor recordings." The "ground truth" here is the expected physiological correlation between blood flow and pulse.
  • Clinical Correlation: In clinical studies, the "ground truth" for the device's intended function was the correlation of measured skull motion with a regular pulse related to the cardiac cycle. This is an observable physiological phenomenon.

• Experts: No specific number or qualifications of "experts" are mentioned for establishing this type of ground truth, as it relies on physical and physiological principles and comparisons to established reference measurements.

4. Adjudication Method for the Test Set

Not applicable in the context of this device's performance evaluation. Adjudication methods (e.g., 2+1, 3+1) are typically used in studies where human experts are making qualitative or subjective assessments that need to be aggregated into a "ground truth" for a diagnostic label. Here, the performance is based on quantifiable physical measurements and their correlation with physiological events.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done

No, an MRMC comparative effectiveness study was not conducted and was not necessary for this device.

• Reasoning: An MRMC study assesses the impact of AI assistance on human reader performance, typically for diagnostic tasks (e.g., radiologists reading images with or without AI). The BrainPulse is a measurement device, not a diagnostic aid that assists a human reader in interpreting complex clinical data. Its output (skull motion data) is intended to be incorporated into a clinician's overall assessment paradigm, but the device itself doesn't offer a diagnostic interpretation or classification that a human "reader" would be evaluating. The submission explicitly states: "Consequently, a demonstration of clinical diagnostic utility in specific patient populations was not required."

6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) was done

Yes, the core of the performance evaluation for BrainPulse, Model 1100, is essentially its standalone (algorithm/hardware-only) performance in measuring cranial motion.

• Measurement Accuracy and Reliability: The bench performance tests (accelerometer stability, repeatability, resolution, PPG accuracy, hardware verification) evaluate the device's ability to accurately and reliably capture the intended data independently.
• Clinical Correlation: The clinical studies confirmed that the device's measurements (algorithm's output) correlated with the cardiac cycle, which is its primary intended function. While clinicians use the data, the device is evaluated on its ability to produce the measurement correctly, not on a human's ability to interpret that measurement for a specific diagnostic outcome.

7. The Type of Ground Truth Used

The "ground truth" for the BrainPulse's evaluation was primarily:

• Physical/Engineering Specifications: For bench testing, this often meant comparing the device's output to known physical inputs or outputs from calibrated reference instruments (e.g., "reference measurement device" for SPL sensor, "accelerometer specifications" for resolution).
• Physiological Correlation: For the PPG sensor, the ground truth was the expected physiological correlation with SpO2 readings (though "visual comparison" suggests a qualitative assessment of this correlation rather than quantitative comparison to a gold standard).
• Observable Physiological Events: In clinical studies, the ground truth for validating the device's intended use was the "regular pulse related to the cardiac cycle," an established physiological event which the device's skull motion measurements were expected to correlate with.
• Absence of Adverse Events: Safety ground truth relied on patient reporting of discomfort or adverse events.

This is distinct from "expert consensus" or "pathology" which are typically ground truths for diagnostic tasks. Outcomes data might be relevant for clinical utility, which was explicitly not assessed.

8. The Sample Size for the Training Set

The document does not mention a training set or any machine learning (ML) or Artificial Intelligence (AI) model that would require a distinct training set. The device appears to be based on direct physical measurements using accelerometers and a PPG sensor. Its "software" is described as managing recordings and saving data, consistent with traditional software, not an ML/AI algorithm that learns from data.

If there were internal algorithms for signal processing or noise reduction, the document does not specify if these were "trained" on data or if they were designed based on known physics and engineering principles. Given the de novo nature from 2014, it's highly likely that any signal processing would be deterministic rather than AI/ML-based.

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

Not applicable, as no training set or specific ML/AI model is described.

Ask a specific question about this device