How 'Micro' is Microperimetry? - Characterizing the Effect of Fundus Tracking on the Psychometric Function

This study demonstrates that while fundus tracking in microperimetry primarily improves psychometric precision by reducing false positives and sharpening sensitivity curves at non-seeing retinal loci, it supports a defensible suprathreshold criterion intensity of 10 to 13 dB for distinguishing seeing from non-seeing retina.

The Gist

Imagine your retina (the back of your eye) is a giant, high-resolution map of a city. Microperimetry is like a drone flying over this city, dropping little "flashlights" (stimuli) at specific addresses to see if the people living there can see the light.

The big question this paper asks is: How steady is the drone's hand?

In the real world, our eyes are never perfectly still. They jitter, drift, and make tiny, involuntary movements (like a shaky hand holding a camera). Microperimetry machines have a feature called "Fundus Tracking," which is like a gimbal on a camera that tries to keep the flashlight perfectly locked on the target, even if your eye moves.

This study wanted to know: Does turning on the "gimbal" (tracking) actually make a difference, or is the drone steady enough on its own? And, if we just want to know "Can you see this light or not?" (a quick check rather than a detailed measurement), what brightness should we use to get the most accurate answer?

Here is the breakdown of their findings using some everyday analogies:

1. The Experiment: The "Blind Spot" Test

The researchers used healthy volunteers and tested two types of locations:

• The "Seeing" Neighborhood: Areas of the retina that work perfectly.
• The "Blind Spot" (The Dark Alley): A natural area in everyone's eye where there are no light sensors (the optic nerve head). It's a guaranteed "dark zone."

They tested these spots with the tracking ON and OFF.

2. The Findings: The "Shaky Hand" Effect

In the "Seeing" Neighborhood:
When the eye was looking at a place that should see the light, the tracking didn't change much. Whether the drone was locked on or just hovering, the people said, "Yes, I see it!"

• Analogy: If you are shining a flashlight on a bright, white wall, it doesn't matter if your hand shakes a little; the wall is still bright.

In the "Blind Spot" (The Dark Alley):
This is where it got interesting. When the drone was supposed to drop a light in the "Dark Alley" (where no one should see anything), the results changed based on the tracking.

• Tracking OFF: The drone's hand shook. Sometimes, the flashlight accidentally drifted out of the dark alley and landed on the "seeing" neighborhood next door. The volunteer would say, "I see it!" even though the light was supposed to be in the blind spot. This is a False Positive.
• Tracking ON: The drone stayed locked. The light stayed in the dark alley. The volunteer correctly said, "I don't see it."
• Analogy: Imagine trying to drop a coin into a tiny hole in a table. If your hand is shaky (no tracking), you might miss the hole and hit the table surface (the seeing retina). If you use a guide rail (tracking), the coin goes straight into the hole.

The "Slope" of the Curve:
The researchers also looked at how "sharp" the transition was between seeing and not seeing.

• Without tracking: The transition was "fuzzy." It was hard to tell exactly where the light stopped being visible.
• With tracking: The transition became "crisp" and steep.
• Analogy: Without tracking, the edge of a shadow looks blurry and gray. With tracking, the edge of the shadow is a sharp, black-and-white line.

3. The Big Question: How Bright Should the Flashlight Be?

In clinical practice, doctors often want to do a quick "defect map" to see if a patient has a blind spot. They don't want to do a long, detailed test. They just want to ask: "Can you see this light?"

To do this, they need to pick a specific brightness (criterion).

• Too dim: You might miss a real problem.
• Too bright: You might think a problem exists when it doesn't.

The researchers ran simulations to find the "Goldilocks" brightness. They found that 13 decibels (dB) was the mathematically perfect brightness to separate "seeing" from "not seeing."

• However, they noted that in real-world diseases (like macular degeneration), eyes are generally a bit weaker. So, they concluded that 10 dB is a safe, conservative, and very good choice for doctors to use in practice. It's close enough to the perfect 13 dB to be accurate, but safe enough to catch real problems in weaker eyes.

The Takeaway

1. Tracking Matters: Even in healthy eyes with good focus, turning on the "tracking" feature makes the test more precise. It stops the "flashlight" from accidentally drifting into the wrong neighborhood, which prevents false alarms.
2. The "Sweet Spot" Brightness: For quick checks to map out blind spots, using a light intensity of 10 to 13 dB is the scientifically backed "sweet spot." It's bright enough to be seen by healthy eyes but dim enough to be invisible to truly damaged areas, giving the most accurate map of the eye's health.

In short: Microperimetry is "micro" enough to be precise, but only if you use the tracking feature to keep your hand steady, and you pick the right brightness to ask the question.

Technical Summary

1. Problem Statement

Microperimetry (fundus-controlled perimetry) is a standard tool for assessing macular function, particularly in diseases like Geographic Atrophy (GA) where fixation is unstable. While clinical "face validity" suggests stimuli are delivered accurately (e.g., no response in atrophic zones), this relies on aggregate data.

• The Gap: It remains unclear how precisely individual stimuli are placed at intended retinal locations, especially in regions with steep sensitivity gradients (junctional zones). Small fixation excursions or placement errors can flip a response from "seen" to "not seen."
• The Question: Does fundus tracking provide stimulus-placement precision beyond natural fixation stability in healthy observers? Furthermore, what is the psychometrically grounded criterion intensity for "defect-mapping" (suprathreshold) microperimetry to reliably separate "seeing" from "non-seeing" retina?

2. Methodology

The study utilized a prospective, controlled experimental design with 25 healthy volunteers (median age 27) using the MAIA2 microperimeter.

• Experimental Design:
  • Test Locations: Five specific loci were selected relative to the optic nerve head (used as a natural, deep scotoma with a sharp boundary):
    • Three loci outside the blind spot (+0.25°, +0.75°, +1.25°).
    • Two loci inside the blind spot (−0.25°, −0.75°).
  • Conditions: Frequency-of-Seeing (FoS) functions were measured in four blocks: two with tracking enabled and two with tracking disabled. The order was randomized, and participants were masked to the condition.
  • Stimuli: A method of constant stimuli was used. After an initial ZEST threshold estimation, 15 presentations per intensity level (7 levels spanning ±3 dB) were delivered per block.
• Data Analysis:
  • Psychometric Modeling: FoS data were fitted using cumulative Gaussian psychometric functions within a Bayesian framework. This estimated:
    • Threshold ($m$): Sensitivity midpoint.
    • Slope ($s$): Steepness of the transition (precision).
    • Guess/Lapse rates: False-positive and false-negative asymptotes.
  • Statistical Modeling: Joint Bayesian multivariate mixed-effects models assessed the impact of tracking on thresholds and slopes. Logistic mixed-effects models analyzed fixation displacement to explain false positives/negatives.
  • Optimal Criterion: Posterior predictive simulations were used to identify the stimulus intensity maximizing the Youden index for distinguishing "seeing" vs. "non-seeing" loci.

3. Key Contributions

• Quantification of Tracking Precision: The study moves beyond aggregate clinical observations to quantify the psychophysical impact of fundus tracking on individual stimulus delivery using the MAIA2 device.
• Psychometric Characterization of Scotoma Boundaries: It utilizes the physiologic blind spot as a controlled model to isolate stimulus-placement errors without the confounding variable of residual photoreceptor function found in disease states like GA.
• Evidence-Based Criterion Selection: It provides a data-driven derivation for the optimal suprathreshold intensity for defect-mapping microperimetry, moving away from heuristic choices (0 dB or 10 dB) toward a psychometrically optimized value.

4. Key Results

• Effect of Tracking on Thresholds:
  • Outside Blind Spot: Tracking had little to no effect on threshold estimates at loci 1–3 (differences were small and statistically uncertain).
  • Inside Blind Spot: Tracking significantly lowered threshold estimates (increased sensitivity to non-seeing) at loci 4 and 5.
    • Locus 4: −1.46 dB difference (95% CrI: −2.30 to −0.62).
    • Locus 5: −1.02 dB difference (95% CrI: −1.94 to −0.08).
  • Interpretation: Without tracking, fixation shifts caused stimuli intended for the "non-seeing" blind spot to land on "seeing" retina, artificially inflating sensitivity estimates.
• Effect of Tracking on Slope (Precision):
  • Tracking was associated with steeper psychometric slopes (sharper transition from seeing to non-seeing) at loci 1–3 (outside the blind spot).
  • This indicates that tracking reduces trial-to-trial spatial jitter, making sensitivity estimates more precise even in healthy retina.
• False Positives/Negatives:
  • False Positives: In the absence of tracking, negative horizontal displacements (moving stimulus toward seeing retina) significantly increased the likelihood of false-positive responses, particularly at the blind spot edge.
  • False Negatives: Positive displacements (toward the optic nerve) without tracking increased false-negative rates at seeing loci.
• Optimal Suprathreshold Criterion:
  • Simulation identified 13 dB as the nominally optimal criterion intensity for separating "seeing" from "non-seeing" retina (Youden index: 0.76).
  • However, a 10 dB criterion yielded comparable performance (Youden index: 0.74) and is considered a conservative, robust choice that accounts for disease-related sensitivity loss (e.g., in AMD).
  • Criteria below 10 dB reduced true-positive detection of defects; criteria above 16 dB increased false positives.

5. Significance and Conclusion

• Validation of Tracking: Even in healthy observers with stable fixation, fundus tracking measurably improves the accuracy of stimulus delivery by reducing false positives at scotoma boundaries and sharpening psychometric functions. This suggests the benefit of tracking is likely even greater in patient populations with unstable fixation.
• Clinical Protocol Optimization: The study provides a strong psychometric rationale for using a 10–13 dB criterion in defect-mapping microperimetry.
  • Using 0 dB (maximum intensity) is overly stringent and prone to false negatives due to the "floor effect" and residual stray light.
  • Using 10–13 dB offers the best balance of sensitivity and specificity for identifying deep scotomas, supporting current clinical practices that favor 10 dB over 0 dB.
• Future Directions: The findings underscore the importance of tracking in high-resolution mapping of lesion borders (junctional zones) in conditions like GA, where sub-degree precision is critical for monitoring disease progression.

In summary, the paper demonstrates that "micro" in microperimetry is contingent on active fundus tracking to achieve true spatial precision, and it establishes a psychometrically derived intensity range (10–13 dB) as the optimal standard for clinical defect mapping.