Evaluation of the “Freiburg Acuity VEP” on Commercial Equipment
- 99 Downloads
To assess the implementation and outcome quality of the Freiburg Acuity VEP methodology (Bach et al. in Br J Ophthalmol 92:396–403, 2008) on the Diagnosys Espion Profile and E3 electrophysiology systems.
We recorded visual evoked potentials (VEPs) from both eyes of 24 participants, where visual acuity (VA) was either full or reduced with scatter foils to approximately 0.5 and 0.8 LogMAR, resulting in a total of 144 recordings. Behavioral VA was measured in each case under the same conditions using the Freiburg Acuity Test (FrACT); VEP-based acuity was assessed with the “heuristic algorithm,” which automatically selects points for regression to zero amplitude.
Behavioral VA ranged from − 0.2 to 1.0 LogMAR. The fully automatic heuristic VEP algorithm resulted in 8 of 144 recordings (6%) that were scored as “no result.” The other 136 recordings (94%) had an outcome of − 0.20 to 1.3 LogMAR (which corresponds to a range of 20/12.5–20/400, or 6/3.8–6/120, in Snellen ratios; or 1.6–0.1 in decimal acuity). The heuristic VEP algorithm agreed with the behavioral VA to within ± 0.31 LogMAR (95% limits of agreement), which is equivalent to approximately three lines on a VA chart.
The successful implementation of the Freiburg Acuity VEP “heuristic algorithm” on a commercial system makes this capability available to a wider group of users. The limits of agreement of ± 0.31 LogMAR are close to the original implementation at the University of Freiburg and we believe are clinically acceptable for a fully automatic, largely objective assessment of visual acuity.
KeywordsVisual acuity Objective VEP Visual evoked potentials Computer
Brief onset checkerboard presentation, yielding relatively high amplitudes,
Temporally in the steady-state region, allowing Fourier transform-based analysis,
Laplace montage, yielding high noise rejection,
Application of the Meigen/Bach statistic , yielding noise-corrected response and significance,
An automated “heuristic algorithm” for regression point selection, enabling an acuity estimate, or the outcome of “no result” even when a “notch”  is present.
This approach was used in the Freiburg laboratory for a decade with high testability, (i.e., a “no result outcome” occurs in only 5–10% of the cases); problems in amblyopia have been described [5, 6], and the algorithm was successfully extended to very low acuities (≈ 2 LogMAR) . The stimulation and recording system was also used in other laboratories and is available free of charge . However, the platform hardware and software is outdated (e.g., MacOS 9), so a re-implementation was needed to enable a broader set of clinical and research users to operate the method in their clinics. Diagnosys expressed an interest in the method and implemented it following the method previously reported , and we herein report the outcome of that implementation.
Equipment and Stimuli
Steady-state VEPs were recorded using a Diagnosys Espion E3 System (Lowell, MA, USA). Checkerboard stimuli were presented in brief onset mode, two frames (33.3 ms) on at 100 cd/m2 and six frames (100 ms) off, corresponding to 7.5 Hz with a stimulus distance of 180 cm and a contrast of 40%. Three sets of check sizes were used, one for the highest VA range (‘‘Range A’’), one for medium VA range (“Range B”) and one for the lowest VA range (‘‘Range C’’). For Range A, the check sizes were 0.37° to 0.05°, for Range B they were 1.19°–0.17° and for Range C they were 4.0°–0.57°. Six check sizes were used in each Range, and cumulatively across all three Ranges there were twelve unique check sizes (i.e., there is overlap of checks sizes between the Ranges). In this study for the three acuity conditions of each participant, Range A and Range B were always used, and depending on the adjusted acuity of the participant with the strongest Bangerter foil, either Range B or Range C was used for the lowest acuity recording. The benefit of having three check size ranges available is that in clinical use this approach will usually keep the test time shorter for a patient. Typically, the clinic has a general understanding of the acuity range the patient is likely to fall within thereby enabling it to choose one of the three shorter protocols. In cases where the clinic does not have that knowledge, they can run the full set of twelve check sizes on a patient.
Freiburg Acuity Test (FrACT)  measurements were taken on a standard PC with a screen size of 58.5 cm (diagonal) using the same Bangerter foils used for the steady-state VEP recordings, also at a distance of 180 cm from the computer.
There were twenty-four participants in the study (14 male and 10 female), with an age range of 19 to 74 years old (mean age was 46.5), and in each case both eyes were tested. Participants were given the choice of either using their habitual eyeglasses during the test or not, and that same condition was used for all tests for that participant. Under these conditions the participants’ visual acuity ranged from approximately 0.6 to better than − 0.15 LogMAR. Each participant completed one set of tests with full vision conditions (defined as participants with their chosen correction, and no Bangerter foil), which was typically 0.30 LogMAR or better. An additional set of tests were then conducted using a Bangerter foil that was intended to reduce the participants’ acuity to approximately 0.4 LogMAR, and a final step with a foil intended to reduce acuity to approximately 1.0 LogMAR. Since every participant was recorded under several conditions, one might think that the “eyes or patients” problem might arise . Given that we are using descriptive, not inferential statistics, this is not a problem.
Response traces were de-trended and subjected to a discrete Fourier transform (DFT). Because care was taken to choose the analysis interval (1066 ms) to be an integer multiple of the stimulation period, there is no overspill in the spectrum  and the noise can be estimated by averaging the magnitudes recorded at the two neighboring frequencies (6.5 and 8.5 Hz). The ‘true’ response magnitude at 7.5 Hz was calculated by non-linearly subtracting the noise from the magnitude measured at 7.5 Hz, and finally a significance for the response at 7.5 Hz was also calculated .
Responses are recorded over six check sizes. Ideally, the stimulation of the various check sizes would be interleaved, but this was not yet implemented on the system in this study. The six check sizes were selected from a range of 0.05° to 4.0°, as appropriate for the expected VA. The responses were processed as described above, resulting in 6 values for the response magnitude plus the associated significances. From these, the heuristic algorithm, starting at small check sizes, selects as many points as possible up to peak response and avoiding a notch  if present. The resulting points are regressed to zero magnitude on a log(spatial frequency) scale, resulting in the value SF0. SF0 is divided by 17.6 (calibration factor, ), yielding a decimal acuity estimate VAdec(VEP). This is converted to LogMAR using the standard formula: VALogMAR = − log10(VAdec). When insufficient points are found or other irregularities occur, a “no result” outcome is flagged, reducing “testability.”
The relationship between behavioral acuity and the VEP-based acuity estimate is quantified in terms of the Bland-Altman limits of agreement (LoA) . Frequently for such a task the correlation coefficient is computed; that is, however, an inappropriate measure because it is normalized by range [12, 15].
Using the Acuity VEP method from , we found a high testability (94%) and a reasonably close agreement of behavioral and VEP acuity estimates (95% limits of agreement of ≈ 3 lines). Behavioral test-retest LoA can be as low as 0.1 LogMAR [16, 17] but will be markedly higher in a clinical population. Thus, the LoA of ± 0.31 for the Acuity VEP seems acceptable and is very close to the one reported earlier . The possibility of outliers should, of course, always be considered when analyzing patient data in clinic.
The current implementation of the Acuity VEP method steps through the six check sizes sequentially starting from the largest one, recording 40 sweeps at each check size. Each step takes approximately 40 s, resulting in a total of about 4 min of recording per eye. In principle, this allows the technician to stop adding steps with finer checks when the amplitude drops. However, this should never be done because insufficient data may be recorded thereby preventing a proper regression and the test could miss a notch region (e.g., the amplitude might rise again for smaller checks). In the present data set, two recordings had “no result” for this reason. In practice, the protocol is created in a stepwise fashion to allow the patient a brief time to rest at a few points during the test.
The VEP protocol had been set up with digital filtering in the range of 5 to 50 Hz. In hindsight, this seems unnecessarily narrow. However, this filtering does not affect the results because the heuristic algorithm is solely based on the 7.5 Hz spectral line (response) and its immediate neighbors (noise estimators). Even some mains intrusion (at 50 or 60 Hz, depending on locality) would not have a detrimental effect. In the future, the Acuity VEP protocol will use standard VEP filtering settings.
We have not addressed test-retest agreement here, since this is implicitly covered by the analysis of agreement between the behavioral visual acuity and the VEP acuity outcome. In future work, it may be of interest to specifically assess the test-retest agreement to analyze the relative variance contribution from interindividual vs. intraindividual sources.
The Freiburg Acuity VEP has previously been found to be of substantial aid in the management of patients with non-organic visual loss. The present study shows that the method has been implemented effectively in a commercial system, enabling its use in a broader set of clinical sites; the method is also being validated for pediatric applications. We would also welcome critical third party assessments without our own conflicts of interest. Finally, if the machine learning approach lives up to its promise , it can be applied to the Freiburg Acuity VEP method potentially extending its value to researchers and clinicians.
We thank our participants for their kind cooperation.
MB and JDF designed the study, JDF took the recordings, MB analysed the data and composed figures, both authors contributed equally to the manuscript.
Compliance with ethical standards
Conflict of interest
Michael Bach served as consultant for Diagnosys. Jeff Farmer is CEO of Diagnosys LLC.
All procedures performed in studies involving human participants were in accordance with the ethical standards of the Medical Center, Faculty of Medicine, University of Freiburg, Germany and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards .
Statement of human rights
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.
Statement on the welfare of animals
This article does not contain any studies with animals performed by any of the authors.
Informed consent was obtained from all individual participants included in the study.
- 4.Bach M, Joost W (1989) VEP vs spatial frequency at high contrast: subjects have either a bimodal or single-peaked response function. In: Kulikowski J, Dickinson C, Murray I (eds) Seeing contour and colour. Pergamon Press, Oxford, pp 478–484Google Scholar
- 8.Bach M (2007) Freiburg evoked potentials. http://www.michaelbach.de/ep2000.html. Accessed 19 Aug 2013
- 11.Fahle M, Bach M (2006) Origin of the visual evoked potentials. In: Heckenlively J, Arden G (eds) Principles and practice of clinical electrophysiology of vision. MIT Press, Cambridge, pp 207–234Google Scholar
- 15.McAlinden C, Khadka J, Pesudovs K (2011) Statistical methods for conducting agreement (comparison of clinical tests) and precision (repeatability or reproducibility) studies in optometry and ophthalmology. Ophthalmic Physiol Opt 31:330–338. https://doi.org/10.1111/j.1475-1313.2011.00851.x CrossRefPubMedGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.