Ahmad Hajreza Jafarabadi1 
, Mir Mohsen Pedram *2 , Mohammadreza Abolghasemi dehaqani3 , Mohammad Abdolrahmani1 , Einollah Pasha2
, Mir Mohsen Pedram *2 , Mohammadreza Abolghasemi dehaqani3 , Mohammad Abdolrahmani1 , Einollah Pasha2
1- ICSS
2- Kharazmi University
3- Tehran University
2- Kharazmi University
3- Tehran University
Abstract: (147 Views)
This study investigates the computational mechanisms underlying binocular disparity discrimination in visual area V4 through a novel modeling approach. We examined the ability of V4 neurons to discriminate between fine disparities near the fixation plane and developed a comprehensive computational framework to characterize this process. Our central hypothesis posits that disparity discrimination ability in V4 depends on two key factors: the absolute difference between disparity pairs and their proximity to zero disparity.
We analyzed electrophysiological data from 156 V4 neurons recorded from macaque monkeys during presentation of random-dot stereograms with varying binocular disparities (±1.2°, ±0.6°, ±0.3°, 0°) and correlation levels. A novel population-level metric, the Disparity Discrimination Ability Index (DDAI), was introduced to quantify the neural population’s capacity to distinguish between disparity pairs using ROC-based analysis. The DDAI was computed as the average normalized area under the ROC curve across all neurons for each stimulus pair.
Our results confirmed that V4 neurons exhibit specialized tuning for near-zero disparities, with response variability (coefficient of variation) being minimal for stimuli close to zero disparity and increasing with absolute disparity magnitude. The computational model DCM (DDAI Computational Model) successfully predicted experimental DDAI patterns with high accuracy (Pearson r = 0.969, Spearman ρ = 0.887). Statistical equivalence testing (TOST) confirmed that model predictions were practically equivalent to experimental data (mean difference = 0.00018, 90% CI within ±0.03).
These findings demonstrate that V4 employs an efficient coding strategy prioritizing precision near the fixation depth while maintaining coarser encoding for larger disparities. The DCM framework provides a robust foundation for understanding population-level disparity processing and offers potential applications in artificial vision systems and clinical assessment of stereoscopic vision disorders.
We analyzed electrophysiological data from 156 V4 neurons recorded from macaque monkeys during presentation of random-dot stereograms with varying binocular disparities (±1.2°, ±0.6°, ±0.3°, 0°) and correlation levels. A novel population-level metric, the Disparity Discrimination Ability Index (DDAI), was introduced to quantify the neural population’s capacity to distinguish between disparity pairs using ROC-based analysis. The DDAI was computed as the average normalized area under the ROC curve across all neurons for each stimulus pair.
Our results confirmed that V4 neurons exhibit specialized tuning for near-zero disparities, with response variability (coefficient of variation) being minimal for stimuli close to zero disparity and increasing with absolute disparity magnitude. The computational model DCM (DDAI Computational Model) successfully predicted experimental DDAI patterns with high accuracy (Pearson r = 0.969, Spearman ρ = 0.887). Statistical equivalence testing (TOST) confirmed that model predictions were practically equivalent to experimental data (mean difference = 0.00018, 90% CI within ±0.03).
These findings demonstrate that V4 employs an efficient coding strategy prioritizing precision near the fixation depth while maintaining coarser encoding for larger disparities. The DCM framework provides a robust foundation for understanding population-level disparity processing and offers potential applications in artificial vision systems and clinical assessment of stereoscopic vision disorders.
Keywords: Binocular Disparity, V4 Area, Computational modeling, Depth Perception, Disparity Discrimination Ability
Type of Study: Research |
Received: 2025/06/8 | Accepted: 2025/07/22
Received: 2025/06/8 | Accepted: 2025/07/22
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |