(25 days)
The Bio-logic Evoked Potential (EP) product family is indicated for use in the recording and analysis of human physiological data necessary for the diagnosis of auditory and hearing-related disorders. An auditory stimulus (click, tone, etc.) is presented to the patient's ear through an earphone or headphoners, and the Brainstem Auditory Evoked Response from the patient is recorded using EEG electrodes placed on the scalp. Although this Brainstem Response is very low in amplitude (with respect to surrounding EEG "noise"), the stimulus-response cycle is repeated many times and the resulting responses are averaged from the time of the stimuli. The random noise averages to zero, but if the Brainstem Response signal is present, it's signal will be easily determined in the averaged signal.
The Bio-logic EP System can be used for patients of all ages, from children to adults, including infants and geriatric patients. It is especially indicated for use in testing individuals for whom behavioral audiometric results are deemed unreliable, such as infants, young children, and cognitively impaired or uncooperative adults. The use of the Bio-logic EP family of products is to be performed under the prescription and supervision of a physician or other trained health care professional.
The primary feature modification represented in this Special 510(k) is for the use of a new algorithm and protocol to generally assist in test data interpretation, and specifically assist in the assessment of signal-tonoise ratio and the quality of the Brainstem Auditory Evoked Response in infants. Based on this automatic assessment, the speed of testing may be reduced and/or the quality of the data recording may be improved. This new feature is used in conjunction with the current EP program, without compromising the quality of recorded data or limiting the control and flexibility of the health care professional administering the test.
The Bio-logic Evoked Potential family of products is intended to be used for the recording and analysis of human physiological data for the purpose of neurological diagnosis and treatment of sensory disorders. The predicate device referenced above is the latest in a series of this type marketed by Bio-logic. Other related devices comprising the Evoked Potential family include:
-
- 510(k) #K803226 Bio-logic Evoked Response Stimulators.
-
- 510(k) #K842543 Bio-logic Evoked Potential System.
-
- 510(k) #K844992 Bio-logic Portable Evoked Response System.
-
- 510(k) #K862690 Bio-logic Traveler LT System.
The predicate device performs Evoked Potential recording and analysis functions, providing up to 8 channels of simultaneous data recording. This device has both hardware and software modifications and unprovements over the related devices. Related Device #1 above is the first Evoked Potential device marketed by Bio-logic. It provides for up to 4 channels of data recording. Related Device #2 above is a hardware/software modification to the first device. Related Device #3 above is similar to #2, but utilizes a "portable" computer for ease of use and transportability. Device #4 utilizes hardware variations over devices #1, #2 and #3, primarily to enhance size-reduction and portability. It provides for a maximum of 2 channels of data recording. Trade names of "Traveler", "Express" and "LT" are associated with these transportable devices. Data recording hardware is available in three variations: the "E" Series, for up to 2 channels of data recording; the "SE" Series, for up to 4 channels of data recording; and the "Explorer" Series, for up to 8 channels of data recording. The Evoked Potential software is essentially the same for all of these products, with variations in models to accommodate differences in the hardware.
Evoked Potential systems can be used for three different kinds of tests: Auditory Evoked Potentials (AEP), Visual Evoked Potentials (VEP), and Somatosensory Evoked Potentials (SEP). These variations are called "modalities", and are offered as options in all three models of hardware marketed by Bio-logic. Each modality has its own unique hardware requirements. The modifications associated with this new modified device are to the software only, and do not change the hardware in any way. Also, the change only affects the operation of the Auditory Evoked Potential software functionality. There are no changes to any part of the VEP or SEP hardware or software.
The AEP test works on the basis of repeating a stimulus-response cycle. An auditory stimulation (click, tone, etc.) is presented to the patient through the use of an earphone or headphones. The EEG response from the brain is read through the use of scalp electrodes placed on the patient. The response time of interest is approximately from 1 - 20 milliseconds following the stimulus. The response voltage for this time period is amplified, digitized and stored in the AEP system computer's memory. The stimulation is then repeated, the EEG response is read again, and this cycle is repeated many times . Each time the response is read, it is averaged together with all previous responses. The final data record is the result of averaging several thousand (usually 2000-3000) responses. This averaging process is necessary because the EEG signal is very small, much lower in voltage than the surrounding EEG "noise" present in the recording. The noise is averaged out over the many readings, because the noise will have an average net value of zero. The result from the averaging process will be the signal.
Some of the EEG responses may have large amounts of noise or other artifact caused by random events such as patient movement or externally-generated electrical noise. These artifacts are usually characterized by very high amplitude voltages (relative to normal EEG levels). The EP program automatically monitors the response for abnormally-high voltage levels. When a response contains such artifacts, the response is discarded, not averaged, and not counted in the cycle count.
In preparation for running an AEP test, the user (EP Technologist, etc.) defines the test parameters through the use of a test protocol. In standard testing situations, most of these parameters will remain the same from test. One of the parameters is the number of stimulation cycles to be used. In order for the test to be completed as quickly as possible, it is desirable to use the smallest number of cycles. However, if the number of cycles is too small, the average will still contain large amounts of "hoise" and the quality of the data may be lowered. In general, the higher the number of cycles, the better will be the quality and reliability of the data. So, there is a trade-off between the time to perform the test and the quality of the test results.
Using the standard AEP program, the normal procedure would be to setup the number of stimulations for some nominal number at the lower end of the range, say 2000. The test would then be run to completion and the data manually reviewed by the user. If the quality of the data is considered to be too low, the test would be re-run with a higher number of stimulations. Another approach is for the user to set the number of stimulations to be a number on the high end of the range and continually monitor the accumulating average that is continuously displayed on the computer's monitor screen. Because the averaged data is updated to the screen every few seconds, it is possible to manually stop the test after the data looks "good enough" to the user. While this is admittedly a somewhat subjective measure not necessarily consistent from one user to the next, it is acceptable because of the training that a registered EP Technologist or Audiologist receives in the interpretation of such data records.
This Special 510(k) is for a modification to the standard Bio-logic AEP device, adding a software algorithm that automatically runs in the "background" while the AEP data recording is progressing. The purpose of this algorithm is to calculate a statistical value called "POVR", standing for "Point-Optimized Variance Ratio". For several (up to 10) strategically-selected data points at specific time latencies from the time of the stimulus, the algorithm makes a calculation intended to represent signal-to-noise ratio (S/N) at these points. This calculation is run on the accumulated data average every 256 stimulation-response cycles ("sweeps"). The result of the calculation is the POVR number, which is a statistical measure of the "power" level of the signal with respect to the noise. In normal testing, the POVR number will gradually increase as the number of sweeps increases and the resulting S/N ratio gets larger. When the POVR number reaches a specific threshold, the quality of the recording is considered to be high, and the test is automatically ended.
There are two additional conditions to the termination of the test recording:
-
- The number of artifact sweeps must be lower than a pre-defined percentage of the sweeps, and
-
- The number of sweeps must be greater than some pre-defined minimal number.
In addition, the maximum number of sweeps is set by the user in the AEP protocol, so that the test recording will also be ended when this maximum is reached. In this case, the POVR calculation will be below the stopping threshold, indicating that the quality of the data may still be too low. A second POVR number, less than the stopping POVR number, is used by the algorithm to make a pass/refer recommendation to the user. In any case, because the standard AEP capabilities still exist in the software, the user may evaluate the data recording and make the determination as to whether or not the recording is of acceptable quality, or if more testing is necessary.
This modification to the standard Bio-logic AEP program allows for a more objective and quantifiable measure of the data recording quality, while also achieving test completion more quickly in many cases. In those cases where the testing time is not reduced, but the POVR threshold is ultimately reached, the resulting data recording quality will likely be higher than if the same test had been performed manually. This is because the manual test is terminated when the number of sweeps reaches the pre-set number in the protocol, regardless of the quality of the result. The modified program with the POVR algorithm will continue the stimulus-response cycle until the POVR threshold is reached.
Below is a description of the acceptance criteria and the study to prove the device meets the acceptance criteria.
1. Table of Acceptance Criteria and Reported Device Performance:
The provided K99 summary does not explicitly define quantitative acceptance criteria for the new POVR algorithm. Instead, it describes qualitative benefits and functional improvements.
| Acceptance Criteria (Inferred from Document) | Reported Device Performance (Summary) |
|---|---|
| Reduced Test Completion Time | "The use of the POVR algorithm will usually reduce time for test completion" |
| Objective Quality Determination | "The use of the POVR algorithm... offers an objective basis for determination of data recording quality." |
| Improved Data Recording Quality | "the quality of the data recording may be improved." (When POVR threshold is ultimately reached) |
| Equivalent Safety Characteristics | "No differences. There is no change to any of the patient-connected hardware or the hardware control software." |
| Equivalent Intended Use | "No differences." |
| Equivalent Patient Population | "No differences. However, the optimal POVR points may differ between infant and adult populations. The present POVR algorithm is optimized for infants." |
2. Sample Size Used for the Test Set and Data Provenance:
The document does not specify a separate "test set" in the context of a clinical trial or formal validation study with a defined sample size. The performance claims regarding reduced test time and improved quality are descriptive based on the algorithm's design and expected behavior, rather than formal statistical testing on a distinct dataset.
- Sample Size: Not specified.
- Data Provenance: Not specified. The document implies internal development and testing rather than external clinical data.
3. Number of Experts Used to Establish Ground Truth for the Test Set and Qualifications:
Given the lack of a formal test set described for a clinical study, there is no mention of experts establishing a "ground truth" for such a set in the traditional sense. The document refers to the role of trained professionals in evaluating data:
- Number of Experts: Not specified.
- Qualifications of Experts: "registered EP Technologist or Audiologist receives in the interpretation of such data records." and "All program 'recommendations' are subject to review by the EP Technologist or Physician."
4. Adjudication Method for the Test Set:
Not applicable, as a formal test set and a related adjudication process are not described in the document. The document highlights that human professionals retain ultimate oversight and decision-making authority: "All program 'recommendations' are subject to review by the EP Technologist or Physician, and may be modified, overridden or deleted as determined by a qualified user."
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study:
No mention of a multi-reader multi-case (MRMC) comparative effectiveness study. The document focuses on the automated nature of the POVR algorithm and its potential benefits (speed and objectivity) compared to purely manual termination of tests. It does not evaluate human reader performance with and without AI assistance.
6. Standalone Performance Study:
The document describes the POVR algorithm as running "in the background" and making calculations to determine when to automatically end the test. This suggests a standalone functional evaluation of the algorithm's performance in achieving its stated goals (e.g., reaching a statistical threshold correctly and terminating a recording). However, a formal "standalone study" with quantifiable metrics presented is not detailed. The description of its operation serves as the primary evidence of its standalone function.
7. Type of Ground Truth Used:
The ground truth for the device's function is the inherent signal-to-noise ratio within the recorded Auditory Evoked Potential (AEP) data. The POVR algorithm's calculation is intended to statistically represent this signal-to-noise ratio. The ultimate "ground truth" for a meaningful AEP recording is the presence of a discernible Brainstem Auditory Evoked Response, which is assessed by trained professionals. The algorithm aims to objectively quantify this.
- Type of Ground Truth: The intrinsic statistical properties of the AEP signal and noise, as interpreted and validated by expert judgment of AEP recordings.
8. Sample Size for the Training Set:
Not specified. The document does not discuss a "training set" in the context of machine learning, but rather describes the design and optimization of a statistical algorithm ("Point-Optimized Variance Ratio"). The optimization for infants is mentioned, suggesting specific parameter tuning, but not based on a quantified training dataset with ground truth labels.
9. How the Ground Truth for the Training Set Was Established:
Not applicable, as a classical machine learning "training set" with ground truth establishment is not discussed. The algorithm appears to be designed based on neurophysiological principles and statistical methods to quantify signal-to-noise ratio in AEPs, rather than being trained on a labeled dataset. The statement "The present POVR algorithm is optimized for infants" suggests internal parameter tuning based on understanding infant AEP characteristics, but the method for establishing "ground truth" during this optimization is not detailed.
§ 882.1900 Evoked response auditory stimulator.
(a)
Identification. An evoked response auditory stimulator is a device that produces a sound stimulus for use in evoked response measurements or electroencephalogram activation.(b)
Classification. Class II (performance standards).