(97 days)
The Delphi Amplifier is intended to be used by or under the direction of a physician for acquisition of EEG signals and to transmit them digitally to a computer. The device is intended for use on humans. The device is intended for use in a clinical environment (e.g., hospital, physician's office, etc.). The device is not intended for use in life support systems.
The Delphi Amplifier has been designed as a mobile recording device for EEG (electroencephalography) signals. It provides access to recorded data over a USB connection to external software over its signal driver interface. The Delphi Amplifier is powered via a USB connection and does not require any additional power source. The Delphi Amplifier supports EEG acquisition and storage by Referential DC input channels, and Parallel trigger input channels. The Delphi Amplifier enables connection of up to 48 referential channels and 8bit trigger input channel. Impedance values can be measured for all referential electrodes as well as the reference. An SDK (Software Development Kit) is available for direct amplifier communication. The device is powered by the USB port connection to a computer. The USB port supports interface to the computer running the SDK.
This document is a 510(k) summary for the Delphi Amplifier, a physiological signal amplifier, specifically for EEG signals. It primarily focuses on demonstrating substantial equivalence to a predicate device (eego amplifiers). This document describes non-clinical bench testing for technical performance and safety, but it does not describe a clinical study involving human patients, human readers, or the establishment of ground truth for a diagnostic AI algorithm. Therefore, many of the requested elements for describing an AI/ML-based diagnostic device's acceptance criteria and study proving its performance cannot be answered from this document.
However, I can extract the information related to the device's technical performance testing:
1. A table of acceptance criteria and the reported device performance
The document does not explicitly present a table of acceptance criteria for each test with numerical performance targets vs. reported values. Instead, it lists the tests performed and states a "Pass" result, indicating that the device met the underlying requirements of these standards.
| Test Category | Test Method Summary | Acceptance Criteria (Implied) | Reported Device Performance |
|---|---|---|---|
| Safety and Essential Performance (Electrical) | Per IEC 80601-2-26 (Electroencephalographs) and IEC 60601-2-26 | Compliance with standard | Pass |
| General Safety and Essential Performance | Per IEC 60601-1 | Compliance with standard | Pass |
| Usability | Per IEC 60601-1-6 | Compliance with standard | Pass |
| Electromagnetic Disturbances | Per IEC 60601-1-2, 60601-2-26 and IEC TR 60601-4-2 | Compliance with standard | Pass |
| System Design Requirement (Data Packet Loss & Long-term Registration) | Stress test assessing data packet loss over 24 hours as well as long term registration test over 24 hours | No errors encountered | No errors encountered. Test passed |
| Reliability Testing | Multiple units tested for reliability over 3160 hours of continuous use. Quality checks performed every 24 hours and main characteristics checked every 1000 hours. | Meets reliability targets | Pass |
| Software and Firmware Testing | Software and firmware testing to ensure device operates per specifications. | Operates per specifications | Pass |
| System Performance Testing (Supplemental to IEC 80601-2-26) | Common-mode rejection ratio, Noise Test and Impedance test | Meets specified parameters | Pass |
2. Sample sized used for the test set and the data provenance (e.g. country of origin of the data, retrospective or prospective)
- Sample Size: The document mentions "Multiple units tested" for reliability. For other tests, specific unit numbers are not provided, but the language implies testing on representative devices. This is not a study on clinical data, but rather on the device hardware/software itself.
- Data Provenance: Not applicable. The testing is bench testing of a physical device, not analysis of patient data.
3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts (e.g. radiologist with 10 years of experience)
- Not applicable. This device is an amplifier for EEG signals, not an AI diagnostic algorithm that requires expert-established ground truth from clinical images or data. The "ground truth" for these tests is defined by the technical specifications and standards (e.g., IEC 60601 series).
4. Adjudication method (e.g. 2+1, 3+1, none) for the test set
- Not applicable. This is not a human assessment study.
5. 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. This is not an AI/ML diagnostic device requiring an MRMC study. It is a physiological signal amplifier.
6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done
- "Standalone performance" in this context would refer to the device's technical specifications and performance in measuring/amplifying EEG signals, which is what the bench tests (e.g., Impedance test, Noise test, CMRR) address. The software and firmware testing also falls under this. The document states these tests were passed.
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc)
- The "ground truth" for the performance of this device is defined by the established technical standards (e.g., IEC 60601 series for medical electrical equipment) and the manufacturer's own design specifications (e.g., "no errors encountered" for data packet loss, "operates per specifications" for software/firmware). It's a hardware/software performance "ground truth" rather than a clinical diagnostic "ground truth."
8. The sample size for the training set
- Not applicable. This is a hardware device (amplifier) with associated software/firmware, not a machine learning model that requires a training set of data.
9. How the ground truth for the training set was established
- Not applicable. As above, this is not an ML model.
§ 882.1835 Physiological signal amplifier.
(a)
Identification. A physiological signal amplifier is a general purpose device used to electrically amplify signals derived from various physiological sources (e.g., the electroencephalogram).(b)
Classification. Class II (performance standards).