1. Neuroscience

Mapping human visual contrast sensitivity and vision loss across the visual field with model-based fMRI

  1. Hugo T Chow-Wing-Bom  Is a corresponding author
  2. Matteo Lisi
  3. Noah C Benson
  4. Freya Lygo-Frett
  5. Patrick Yu-Wai-Man
  6. Frederic Dick
  7. Roni O Maimon-Mor
  8. Tessa M Dekker
  1. Institute of Ophthalmology, University College London (UCL), United Kingdom
  2. Birkbeck/UCL Centre for NeuroImaging, United Kingdom
  3. Department of Psychology, Royal Holloway, University of London, United Kingdom
  4. eScience Institute, University of Washington, United States
  5. Institute of Cognitive Neuroscience, UCL, United Kingdom
  6. John van Geest Centre for Brain Repair and MRC Mitochondrial Biology Unit, Department of Clinical Neurosciences, University of Cambridge, United Kingdom
  7. Cambridge Eye Unit, Addenbrooke’s Hospital, Cambridge University Hospitals NHS Foundation Trust, United Kingdom
  8. Moorfields Eye Hospital NHS Foundation Trust, United Kingdom
  9. Department of Experimental Psychology, UCL, United Kingdom
https://doi.org/10.7554/eLife.105930.3
Share this article
Cite this article
  1. Hugo T Chow-Wing-Bom
  2. Matteo Lisi
  3. Noah C Benson
  4. Freya Lygo-Frett
  5. Patrick Yu-Wai-Man
  6. Frederic Dick
  7. Roni O Maimon-Mor
  8. Tessa M Dekker
(2026)
Mapping human visual contrast sensitivity and vision loss across the visual field with model-based fMRI
eLife 14:RP105930.
https://doi.org/10.7554/eLife.105930.3
  2. Download BibTeX
  3. Download .RIS
17 figures, 4 tables and 1 additional file

Figures

Figure 1
Download asset Open asset
Expected changes in cortical contrast sensitivity in typically sighted controls for two different spatial frequencies (low spatial frequency, 0.3 cycles per degree (cpd), and high spatial frequency, 3.0 cpd), (A) as a function of eccentricity and (B) visual field quadrants.

With reference to eccentricity, we expect higher sensitivity of V1 neuronal populations to (1) lower spatial frequencies in the periphery and (2) higher spatial frequencies in the center of the visual field. By comparison, we expect both high and low spatial frequency stimulation to evoke greatest sensitivity in V1 neuronal populations encoding the left and right visual quadrants, less in the lower, and least in the upper quadrant.

Figure 2
Download asset Open asset
Spatial frequency preference V1 across the visual field.

For each participant (n = 7), five eccentricity bins were defined using subject-specific pRF estimates: 0.5°–2.5°, 2.5°–4.5°, 4.5°–9.5°, 9.5°–15°, and 15°–20°. (A) Average β-values versus stimulus contrasts for low (0.3 cpd; blue) and high (3 cpd; red) spatial frequencies across eccentricity bins. Error bars are the 95% confidence intervals. Thick lines represent the group-level model fit (n=7), thin lines are the individual fits. (B) Slopes projected into visual space for low (0.3 cpd) and high (3 cpd) spatial frequencies. The color scale corresponds to slope estimates, with higher values indicating higher cortical contrast sensitivities. (C) Heat plots for each participant (C1–C7).

Figure 3
Download asset Open asset
Anisotropies in V1 across visual field quadrants for low (0.3 cpd; top row) and high (3 cpd; bottom row) spatial frequencies.

For each participant (n=7), pRF estimates were used to link brain responses to visual field positions. Visual quadrants are ±45° around the cardinal meridians, with slope values for eccentricities between 0.5° and 20°. Horizontal quadrants are left and right; vertical quadrants are upper and lower. (A) Average β-values versus contrast levels for each anisotropy (horizontal versus vertical, upper versus lower, left versus right). Error bars are 95% confidence intervals. Thick lines correspond to group-level model fit, while thin lines are individual fits. (B) Slope values projected onto visual field quadrants. The color scale represents slope estimates, with higher values indicating greater V1 contrast sensitivity. (C) Heat plots for each participant (C1–C7).

Figure 4
Download asset Open asset
Reliability of V1 slopes across sessions, in four of the seven controls.

(A) V1 contrast sensitivity (i.e., slopes) plotted in visual field space (using pRF estimates), showing a consistent session-independent increase in cortical sensitivity toward peripheral locations for the 0.3 cpd condition. Higher slopes (warmer colors) indicate higher V1 sensitivity. (B) Slopes for the 0.3 cpd condition across two visits. Spearman correlation coefficients (ρ) between visits are shown on the figure. (C) Slopes for the 0.3 and 3 cpd conditions collected within the same session. Spearman correlation coefficients (ρ) between visits are shown on the figure.

Figure 5
Download asset Open asset
Effect of eye movements on slope estimation, in participant C5.

Each column represents a distinct eye movement condition: (A) central fixation, (B) 2° eye motion, (C) 5° eye motion, and (D) random eye motion. pRF estimates were used to relate brain responses to visual field positions. Top row: Heatmaps of slope distributions for low (0.3 cycles per degree, cpd) and high (3 cpd) spatial frequency stimuli, with warmer colors indicating steeper slopes. Middle row: Slope values scattered across eccentricity bins (0.5°–2.5°, 2.5°–4.5°, 4.5°–9.5°, 9.5°–15°, and 15°–20°), illustrating how cortical sensitivity varies across the visual field for low (blue) and high (red) spatial frequencies. Error bars represent 95% confidence interval. Bottom row: Fixation stability plots from one out of three runs, including fixation points (blue) and BCEA ellipses (red). BCEA values quantify gaze dispersion: Central Fixation (0.91 deg²), 2° eye motion (2.43 deg²), 5° eye motion (18.41 deg²), and random eye motion (130.79 deg²). Despite increased eye movements and reduced slope values – particularly for high spatial frequency stimuli – the overall pattern of cortical sensitivity remains consistent across conditions.

Figure 6
Download asset Open asset
Effect of eye movements on slope estimation, in participant C6.

Each column represents a distinct eye movement condition: (A) central fixation and (B) 2° eye motion. pRF estimates were used to relate brain responses to visual field positions. Top row: Heatmaps of slope distributions for low (0.3 cycles per degree, cpd) and high (3 cpd) spatial frequency stimuli, with warmer colors indicating steeper slopes. Bottom row: Fixation stability plots from one out of 3 runs, including fixation points (blue) and BCEA ellipses (red). BCEA values quantify gaze dispersion: Central Fixation (0.76 deg²) and 2° eye motion (1.23 deg2).

Figure 7
Download asset Open asset
V1 sensitivity (i.e., slope) to 3 cpd condition, in the region of the artificial scotoma (3°–8°) in participant C5.

(A) Masked region between 3° and 8°, representing the artificial scotoma. (B) Averaged β-values between 0.5° and 20° eccentricities in V1, under normal and masked conditions. 20 eccentricity bins of 1° were selected based on pRF estimates. The color scale indicates the presented contrast levels, and the black dotted lines mark the boundaries of the artificial scotoma. Shaded area represents 95% confidence interval. (C) Average β-values versus contrast levels in V1 neurons encoding the region of artificial scotoma (3°–8°), under normal (blue) and masked (red) conditions. Lines correspond to the model fitted to the data. Error bars (masked by dots) represent 95% confidence interval. (D) Slopes projected back onto the visual field under normal and masked conditions. pRF estimates were used to relate brain responses to visual field locations and to create the 1° eccentricity bins. The boundaries of the simulated scotoma region are represented by dotted lines. The color scale indicates slope estimates, with higher values (warmer colors) corresponding to higher V1 contrast sensitivity. (E) Slopes plotted back into the visual space using the calibrated Benson template instead of pRF estimates.

Figure 8
Download asset Open asset
V1 sensitivity (i.e., slope) to 0.3 cpd condition, in the region of the artificial scotoma (upper right quadrant) in participant C5.

(A) Masked upper right quadrant region, representing the artificial scotoma. (B) Average β-values versus contrast levels in V1 neurons encoding each of the four quadrants: upper left (yellow), upper right (cyan), bottom left (red), and bottom right (blue). Lines correspond to the model fitted to the data. Error bars represent 95% confidence interval. (C) Slopes projected back onto the visual field under normal and masked conditions. pRF estimates were used to relate brain responses to visual field locations and to create the 1° eccentricity bins. The color scale indicates slope estimates, with higher values (warmer colors) corresponding to higher V1 contrast sensitivity. (D) Slopes plotted back into the visual space using the calibrated Benson template instead of pRF estimates.

Figure 9
Download asset Open asset
Correspondence between visual field perimetry map and V1 sensitivity map, in a patient with Leber Hereditary Optic Neuropathy.

(A) Left eye. (B) Right eye. Top row: Gray scale visual field map obtained with the Compass fundus perimeter (CenterVue, Padova, Italy), using Standard Perimetry display convention. Sensitivity is represented with symbols and related dB intervals, with larger values describing better sensitivity. Red circle indicates the area of the visual space stimulated in the fMRI contrast sensitivity task. Middle row: Heatmap of V1 sensitivity (i.e., slopes) across the visual field. Here, we divided the space between the upper-left, upper-right, lower-left, and lower-right quadrants (as opposed to the left, right, upper, and lower quadrants used in previous sections), to match the layout of visual field perimetry maps. We also only show the responses to low spatial frequency condition (0.3 cpd) for visualization. Slope values were averaged for each visual quadrant and 1° eccentricity bin and projected back into the visual space, using the subject-specific pRF map. Color scale corresponds to the steepness of the slope estimate. For better visualization of sensitivity differences across quadrants, we also averaged the slopes between 0.5° and 20° eccentricities in each quadrant, generating a single slope value for each visual quadrant (inset heat plots). The range of slope values was reduced on the color scale for these inset plots. Bottom row: Heatmap of V1 sensitivity (i.e., slopes) across the visual field in response to the 0.3 cpd condition, using the Benson retinotopic template map instead of the pRF data.

Figure 10
Download asset Open asset
Estimation of cortical contrast sensitivity across the visual field.

Participants performed two fMRI tasks in the scanner: a contrast sensitivity task and a pRF mapping task. For each individual and tested spatial frequency, the contribution of each stimulus contrast level to the BOLD signal in the contrast sensitivity task was estimated using a general linear model (GLM) approach. Resulting β-coefficients for each contrast level were then projected onto the individual’s cortical surface. For each vertex in V1, a square root function (R(C)=a∗C) was then fitted to these data, taking the estimated slope a as a measure of cortical contrast sensitivity. V1 slopes were finally averaged across vertices based on their eccentricity and polar angle preference as estimated with pRF mapping, and across participants. Mean slopes were then projected back into the visual space, producing a heatmap projection of V1 contrast sensitivity across eccentricities and along visual field quadrants.

Figure 11
Download asset Open asset
Eccentricity alignment and surface area estimation in V1.

(A) For each hemisphere, the scale parameter in Horton and Hoyt’s linear magnification model Mlinear=scaleE+0.75 was estimated from individual V1 surface area and the 0°–90° eccentricity range. Using this, we derived a cumulative surface-area function f(r), which gives the total cortical area of V1 representing eccentricities from 0° to r. For visualization, the histograms show the differential form of this function – that is, the surface-area density as a function of eccentricity. This corresponds to the derivative dfdr, which reflects the areal magnification (denoted m∗(r)) integrated over polar angle. This representation provides a clearer view of how cortical surface area is distributed across eccentricities while remaining consistent with the underlying cumulative function f(r). Adjusted eccentricity values were obtained by iteratively matching the empirical surface-area distribution to the model using a binary search procedure. (B) Comparison of eccentricity maps generated using the Benson template, the adjusted eccentricity values, and population receptive field (pRF) estimates, showing improved alignment with empirical data.

Appendix 1—figure 1
Download asset Open asset
Schematic of large-field setup.

(A) The participant lies on the scanner bed in the 64-channel coil without top to reduce obstruction and views the 40° screen via a mirror above their face (1). Stimuli are displayed using back-projection from a projector (2) outside the room, onto a screen (3) positioned behind the participant’s head. The top of the screen follows the scanner bore curvature to maximize field of view. Monocular eye-tracking is achieved by mounting the illuminator and camera of the eye-tracker (4) vertically on a support (5) at the back of the scanner bore. An image of the eye is obtained via a dual-mirror setup, including a small mirror inside an aperture cut on the side of the screen (6) and the participant mirror (1). If the right eye is being tracked, the eye-tracker is placed on the left side of the scanner bore, and the eye-tracker mirror (6) is placed in the right screen aperture. (B) Schematic of the large-field screen, with the area of stimulus presentation shown in cyan and the mirror used for eye-tracking in magenta. The mirror is positioned within one of the two rectangular apertures cut on either side of the screen.

Appendix 1—figure 2
Download asset Open asset
V1 contrast sensitivity (i.e., slopes) across five eccentricity bins (0.5°–2.5°, 2.5°–4.5°, 4.5°–9.5°, 9.5°–15°, and 15–20°), defined using the calibrated Benson atlas.

(A) β-values versus contrast levels for each eccentricity bin. Blue and red lines represent the contrast sensitivity model fits for the 0.3 and 3cpd conditions, respectively. Thick lines represent the group-level model fit (n=7), thin lines correspond to individual fits. (B) V1 contrast sensitivity index (i.e., slope) projected back into the visual space using the calibrated Benson atlas, for each eccentricity bin and spatial frequency condition. The color scale indicates slope estimates, with higher values representing higher V1 contrast sensitivity.

Appendix 1—figure 3
Download asset Open asset
V1 contrast sensitivity (i.e., slopes) across visual field quadrants using the calibrated Benson atlas, for the 0.3 and 3 cpd conditions.

Visual quadrants were defined as ±45° regions around the cardinal meridians, and slope values were selected for eccentricities between 0.5° and 20°. Horizontal quadrants are left and right, while upper and lower quadrants define the vertical quadrants. (A) β-values versus contrast levels for each anisotropy effect: horizontal versus vertical, upper versus lower, and left versus right. Thick lines represent group-level model fits (n=7), thin lines correspond to individual fits. Error bars represent the 95% confidence intervals. (B) V1 contrast sensitivity index (i.e., slope) projected back into the upper, lower, left, and right visual field quadrants for the 0.3 and 3 cpd conditions. The color scale represents slope estimates, with higher values indicating higher V1 contrast sensitivity.

Appendix 1—figure 4
Download asset Open asset
Effect of correcting eccentricity distribution in retinotopic templates on cortical contrast sensitivity maps.

Heatmaps show slope values back projected into visual space, with each concentric circle representing a 1° eccentricity bin. (A) Back projection was performed using subject-specific pRF map, (B) Using Benson atlas with eccentricity distribution corrected using the H&H model. (C) Using Benson atlas without correction (original template). Color scale indicates slope values reflecting cortical contrast sensitivity (brighter = larger slopes).

Author response image 1
Download asset Open asset
Author response image 2
Download asset Open asset

Tables

Appendix 1—table 1
Results from the multilevel modeling approach in each individual participant.

pRF mapping was used to link brain responses to visual field locations.

SubIDEffectsDfSum SqMean SqF-valuePr(>F)Signif.
C1SF13720372072.49<2e−16***
Eccentricity135,00335,003681.995<2e−16***
PolarLocation311,258375373.115<2e−16***
SF:Eccentricity123,91423,914465.938<2e−16***
SF:PolarLocation3154651510.0411.32E−06***
Eccentricity:PolarLocation3200671.2970.274
SF:Eccentricity:PolarLocation311583867.5195.04E−05***
Residuals12,378635,29051
C2SF123,70423,704743.628<2e−16***
Eccentricity113,50313,503423.609<2e−16***
PolarLocation346,44215,481485.647<2e−16***
SF:Eccentricity110,58710,587332.135<2e−16***
SF:PolarLocation3298199431.169<2e−16***
Eccentricity:PolarLocation3288196030.126<2e−16***
SF:Eccentricity:PolarLocation33731243.8998.52E−03**
Residuals11,604369,89732
C3SF127271.3740.2411
Eccentricity157857829.0197.33E−08***
PolarLocation373152438122.52<2e−16***
SF:Eccentricity114,71914,719739.59<2e−16***
SF:PolarLocation3257864.3114.80E−03**
Eccentricity:PolarLocation3222574237.261<2e−16***
SF:Eccentricity:PolarLocation35101708.5431.16E−05***
Residuals9860196,22520
C4SF125,29025,290521.73<2e−16***
Eccentricity13970397081.89<2e−16***
PolarLocation334,47511,492237.08<2e−16***
SF:Eccentricity175,05475,0541548.38<2e−16***
SF:PolarLocation3251483817.293.40E−11***
Eccentricity:PolarLocation316,4175472112.89<2e−16***
SF:Eccentricity:PolarLocation33921130726.97<2e−16***
Residuals9896479,68848
C5SF155455416.3325.37E−05***
Eccentricity167176717198.161<2e−16***
PolarLocation38862958.7119.12E−06***
SF:Eccentricity113,43613,436396.384<2e−16***
SF:PolarLocation34753158446.744<2e−16***
Eccentricity:PolarLocation33023100829.727<2e−16***
SF:Eccentricity:PolarLocation3263887925.939<2e−16***
Residuals7728261,95634
C6SF160086008246.099<2e−16***
Eccentricity121,37121,371875.448<2e−16***
PolarLocation311,8503950161.806<2e−16***
SF:Eccentricity112,32712,327504.98<2e−16***
SF:PolarLocation37142389.7452.04E−06***
Eccentricity:PolarLocation3156952321.4257.97E−14***
SF:Eccentricity:PolarLocation376225410.4097.80E−07***
Residuals9360228,49324
C7SF110,70310,703278.799<2e−16***
Eccentricity119,43119,431506.149<2e−16***
PolarLocation318,7546251162.84<2e−16***
SF:Eccentricity119,61119,611510.822<2e−16***
SF:PolarLocation3153511.3252.64E−01
Eccentricity:PolarLocation312,0804027104.885<2e−16***
SF:Eccentricity:PolarLocation3138046011.9837.90E−08***
Residuals10,794414,38438
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘ 1.
Appendix 1—table 2
Post hoc test on ANOVA visual quadrant position effect (PolarLocation) at the individual level.

pRF mapping was used to relate brain responses to visual field locations.

SubIDEffectscontrastestimateSEdft.ratiop.value
C1PolarLocationHM > VM5.51381560.327844412,35416.8183918.917786e−63***
LVM > UVM0.59991870.279605312,3542.1455910.03192504*
LHM > RHM0.70582170.171180712,3544.1232543.759997e−05***
C2PolarLocationHM > VM5.5337510.228459511,58024.222031.771783e−126***
LVM > UVM4.3875440.185297411,58023.678394.420105e−121***
LHM > RHM2.2886930.133636111,58017.126305.942464e−65***
C3PolarLocationHM > VM3.5680240.2172194983616.4258967.786414e−60***
LVM > UVM2.3178750.1801032983612.8697041.339442e−37***
LHM >RHM0.1594110.121437798361.3126970.1893155
C4PolarLocationHM > VM7.34080100.3070954987223.9039728.223761e−123***
LVM > UVM3.15145980.2492052987212.6460432.250910e−36***
LHM > RHM–0.94175190.17945589872–5.2478211.571031e−07***
C5PolarLocationHM > VM2.29974600.287403377048.0018071.405261e-15***
LVM > UVM0.27363110.229627977041.1916280.2334437
LHM > RHM–0.22017300.17283447704–1.2738960.2027387
C6PolarLocationHM > VM1.80400080.225900793367.9858131.561107e−15***
LVM > UVM3.20064110.1818476933617.6006783.057311e−68***
LHM > RHM0.30460820.134024593362.2727790.02306211*
C7PolarLocationHM >VM5.90854930.306762310,77019.2610052.637528e−81***
LVM > UVM2.03341640.261771010,7707.7679228.703410e−15***
LHM > RHM–0.33704170.159934510,770–2.1073730.03510831*
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1.
Appendix 1—table 3
Group-level ANOVA analysis on slope estimates, as function of spatial frequency, eccentricities, and visual quadrant positions.

The Benson template retinotopic map was used to relate brain responses to visual field locations.

Measure of Interest: Slopenum Dfden DfMSEFgesPr(>F)
SF1.00006.000031.4360.61170.0108870.4639018
ECC2.165012.990132.6944.70200.1600270.0270342*
PolarLocation1.772910.637427.3167.03570.1632090.0129521*
SF:Eccentricity2.270613.62387.43829.48220.2218077.552e−06***
SF:PolarLocation1.58859.53095.1732.10250.0097930.1783533
Eccentricity: PolarLocation12.000072.00005.2532.96450.0966340.0020990**
SF: Eccentricity: PolarLocation12.000072.00001.1953.52120.0280910.0003995***
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1.
Appendix 1—table 4
Post hoc t-test analysis for the visual quadrant position effect (PolarLocation) and the interaction between spatial frequency and eccentricities (SF:Eccentricity) reported in Appendix 1—table 3.

The Benson template retinotopic map was used to relate brain responses to visual field locations.

Measure of Interest: SlopecontrastestimateSEdft.ratiop.value
PolarLocationHorizontal > Vertical4.33782401.210175263.58445960.01158079
Lower > Upper0.52495200.705373660.74421840.48485944
Left > Right–0.22651940.39226386–0.57746710.58462631
SF:Eccentricity3–0.3 cpd: 0.5°–2.5°4.59176190.796084665.76793200.001185113
3–0.3 cpd: 2.5°–4.5°2.58012431.019274562.53133400.044599322
3–0.3 cpd: 4.5°–9.5°–0.29276350.99818126–0.29329690.779175269
3–0.3 cpd: 9.5°–15°–1.97827890.60427276–3.27381820.016953076
3–0.3 cpd: 15°–20°–2.28030800.64626586–3.52843680.012391168

Additional files

MDAR checklist
https://cdn.elifesciences.org/articles/105930/elife-105930-mdarchecklist1-v1.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Hugo T Chow-Wing-Bom
  2. Matteo Lisi
  3. Noah C Benson
  4. Freya Lygo-Frett
  5. Patrick Yu-Wai-Man
  6. Frederic Dick
  7. Roni O Maimon-Mor
  8. Tessa M Dekker
(2026)
Mapping human visual contrast sensitivity and vision loss across the visual field with model-based fMRI
eLife 14:RP105930.
https://doi.org/10.7554/eLife.105930.3