Development of a transformation model to analyze horizontal saccades using electrooculography through correlation between video-oculography and electrooculography

This study develops and validates a mathematical transformation model that converts electrooculography (EOG) data into video-oculography (VOG)-equivalent values for horizontal saccades, demonstrating that optimized EOG filtering can serve as a cost-effective, high-precision alternative to the gold-standard VOG system.

The Gist

Imagine your eyes are like tiny, powerful searchlights. When you quickly snap your gaze from one object to another, that rapid jump is called a saccade. Doctors love studying these jumps because they are like a "report card" for your brain's wiring. If the jumps are too slow, too fast, or shaky, it can tell a doctor that something is wrong with your brain or nerves.

Currently, the "gold standard" for measuring these jumps is a high-tech camera system called VOG (Video-Oculography). Think of VOG as a high-definition security camera that watches your eyes 24/7. It's incredibly accurate, but it's also expensive, bulky, and requires you to sit perfectly still in front of a screen. It's like trying to take a high-quality photo with a massive studio camera when you just want a quick snapshot.

The Problem:
What if you are in a hospital bed, asleep, or too sick to sit up? The big camera can't see your eyes, or you can't cooperate with it. We need a cheaper, smaller, and more portable way to measure these eye jumps.

The Solution:
Enter EOG (Electrooculography). This is an old-school technique used mostly for sleep studies. Instead of a camera, it uses little stickers (electrodes) placed around your eye. It doesn't "see" the eye; it "feels" the electricity.

Here's the cool part: Your eye acts like a tiny battery. The front (cornea) is positive, and the back is negative. When you look left, the positive side moves closer to the left sticker, and the voltage changes. When you look right, it moves away. It's like a magnetic compass that changes its reading based on which way the needle (your eye) points.

The Challenge:
The problem is that the camera (VOG) measures degrees (how far you looked), while the stickers (EOG) measure voltage (how much electricity changed). They speak different languages. You can't just compare a "degree" to a "volt" directly.

The Breakthrough:
This paper is like a translator that finally lets these two languages talk to each other.

1. The Experiment: The researchers put both the camera and the stickers on four healthy volunteers. They asked them to look left and right at specific targets (fixed) and then at random targets (random).
2. The Math: They built a mathematical "dictionary" (a transformation model). They figured out exactly how to convert the "voltage jump" from the stickers into the "degree jump" that the camera sees.
   • Analogy: Imagine you have a recipe that tells you how many cups of flour make a cake. But you only have a scale that weighs the flour in grams. This study created the formula to say, "Okay, 200 grams of flour equals exactly 1 cup."
3. The Filter: Just like tuning a radio to get rid of static, they had to find the perfect "filter" settings for the electrical signal. They found that a specific setting (0.3 Hz high-pass and 35 Hz low-pass) made the signal crystal clear, removing the "static" (noise) without distorting the "music" (the eye movement).

The Results:
When they used their new "translator" on the EOG data, the results matched the expensive camera almost perfectly!

• Correlation: The match was so strong (95% for rightward looks, 93% for leftward) that it's like looking in a mirror.
• Validation: They tested it on random eye movements, and it still worked.

Why This Matters:
This is a game-changer for hospitals and research.

• Portability: You can strap these little stickers on a patient in the ICU, in an ambulance, or even while they are sleeping. You don't need a giant camera setup.
• Cost: EOG equipment is cheap compared to the high-tech cameras.
• Accessibility: It allows doctors to check brain function in patients who are comatose, have droopy eyelids, or can't sit up—situations where the camera fails.

In a Nutshell:
The researchers took a low-tech, cheap method (electrodes) and taught it to speak the high-tech language of the gold-standard camera. They proved that with the right math and a little signal cleaning, you can get professional-grade eye movement data without needing a $50,000 camera system. It's like turning a simple walkie-talkie into a crystal-clear video call just by improving the signal processing.

Technical Summary

1. Problem Statement

• Clinical Limitation of Gold Standard: Video-oculography (VOG) is the established gold standard for analyzing saccadic eye movements (rapid shifts in gaze). However, VOG systems are expensive, bulky, and lack portability. They require stable visual access to the pupils and significant patient cooperation, making them unsuitable for bedridden patients, those in intensive care units (ICUs), or individuals with eyelid abnormalities (e.g., ptosis) or altered mental status.
• Untapped Potential of EOG: Electrooculography (EOG) detects cornea-retinal potential changes and is widely used in sleep studies (polysomnography). While cost-effective and portable, EOG has historically been limited to qualitative assessments (e.g., detecting eye opening/closing) rather than quantitative saccadic analysis.
• Research Gaps:
  1. There is insufficient validation of the quantitative relationship between EOG-derived and VOG-derived saccadic parameters.
  2. The impact of signal filtering (specifically High-Pass Filter cut-off frequencies) on EOG waveform morphology and velocity estimation has not been systematically optimized for VOG-equivalent conversion.

2. Methodology

The study employed a prospective observational design involving 4 healthy adults (ages 33–44) with no neurological or sleep disorders.

• Data Acquisition:

  • Simultaneous Recording: Horizontal eye movements were recorded simultaneously using VOG (SLVNG, SLMED) and EOG (Comet-PLUS PSG system).
  • Tasks: Subjects performed two types of tasks:
    1. Fixed Saccades: Gaze shifts to targets at ±20° (Derivation Dataset).
    2. Random Saccades: Gaze shifts to targets appearing randomly within ±30° (Validation Dataset).
  • Sampling Rates: EOG at 200 Hz; VOG at 120 Hz.
  • Synchronization: Time alignment was achieved by matching the onset of visual red-dot cues in both systems.

• Signal Processing & Filtering:

  • Filtering Strategy: The study systematically evaluated the impact of High-Pass Filters (HPF) and Low-Pass Filters (LPF).
  • HPF Evaluation: Synthetic and real EOG signals were tested with HPF cut-offs at 0.1, 0.3, and 1 Hz to assess waveform distortion and peak velocity stability.
  • LPF Selection: Various LPF cut-offs were tested, with 35 Hz identified as optimal for consistency.
  • Optimal Configuration: The final model utilized 0.3 Hz HPF and 35 Hz LPF.

• Mathematical Transformation Model:

  • The authors derived a mathematical relationship based on the physics of the cornea-retinal dipole.
  • Core Logic: EOG voltage ($V$) is inversely proportional to the square of the distance ($d$) between the cornea and the electrode. As the eye rotates by angle $\theta$, the cornea displaces, changing $d$ and thus $V$.
  • Derivation: By relating the rate of voltage change ($\Delta V / \Delta t$) to the angular velocity ($\Delta \theta / \Delta t$), a linear transformation model was established:
    $$V_{saccade} \approx C \times EOG_{velocity}$$
  • Directional Constants: Separate proportional constants ($C$) were derived for rightward and leftward saccades to account for geometric asymmetry relative to the electrode placement.

• Statistical Analysis:

  • Uncentered Pearson Correlation: Used to measure the proportional relationship between EOG and VOG velocities (forcing the line through the origin).
  • Paired t-tests: Used to compare VOG velocities against converted EOG velocities to check for statistical significance.

3. Key Contributions

1. Mathematical Transformation Model: The study successfully developed and validated a formula to convert raw EOG voltage change rates into VOG-equivalent peak saccadic velocities (°/s).
2. Filter Optimization: It systematically identified 0.3 Hz HPF and 35 Hz LPF as the optimal filtering parameters to minimize waveform distortion while maintaining accurate velocity estimation.
3. Directional Calibration: The model accounts for directional asymmetry, providing distinct conversion constants for rightward ($C \approx 0.1085$) and leftward ($C \approx 0.1595$) saccades, addressing the non-linear effects of electrode placement.
4. Validation Framework: The model was rigorously tested using a derivation dataset (fixed saccades) and a separate validation dataset (random saccades), proving its generalizability.

4. Results

• Correlation Strength:
  • Rightward Saccades: Strong correlation ($r = 0.95$, $p < 0.0001$).
  • Leftward Saccades: Strong correlation ($r = 0.93$, $p < 0.0001$).
• Model Accuracy:
  • After applying the transformation model, there were no statistically significant differences between the converted EOG velocities and the actual VOG measurements in both the derivation and validation datasets.
  • The model successfully predicted VOG-equivalent values for random saccades using constants derived from fixed saccades.
• Filtering Impact:
  • Higher HPF cut-offs (e.g., 1 Hz) caused progressive waveform distortion and underestimation of peak velocities.
  • The 0.3 Hz setting provided the best balance of signal fidelity and noise reduction.
• Asymmetry Observation:
  • Raw EOG data showed rightward movements appearing faster than leftward ones (due to the electrode being placed on the right side, closer to the cornea during rightward gaze).
  • The transformation model successfully corrected this physical asymmetry, aligning the converted EOG data with VOG norms.

5. Significance and Implications

• Clinical Accessibility: This research validates EOG as a viable, low-cost, and portable alternative to VOG. It enables quantitative saccadic analysis in settings where VOG is impossible, such as:
  • Intensive Care Units (ICUs) for comatose or sedated patients.
  • Bedridden patients who cannot sit upright or cooperate with video tracking.
  • Patients with ptosis or severe eyelid abnormalities.
• Integration with Existing Workflows: Since EOG is standard in polysomnography (sleep studies) and EEG, this model allows for the extraction of neurological biomarkers (saccadic velocity) from existing sleep or neuro-monitoring data without additional hardware.
• Neurological Diagnostics: Saccadic velocity is a critical biomarker for various neurological and psychiatric conditions (e.g., Parkinson's, schizophrenia, brainstem lesions). This study expands the diagnostic reach of these biomarkers to non-ambulatory populations.
• Future Directions: While the study was limited to 4 healthy subjects and horizontal saccades, it lays the groundwork for future validation in larger populations and patients with neurological disorders, as well as the application of the model to vertical saccades and complex eye-tracking tasks.

Conclusion: The study provides a robust, mathematically derived framework that bridges the gap between low-fidelity EOG and high-precision VOG, effectively democratizing access to quantitative saccadic analysis for clinical and research applications.