Framework to Simulate Perceived Images Affected by Human Visual System Disorders.

Area of Science:

• Computational neuroscience
• Computer vision
• Human visual system modeling

Background:

• Human visual system (HVS) disorders, including color vision deficiencies and traumatic brain injury (TBI)-related impairments, significantly alter visual perception.
• Existing models often lack the flexibility to simulate diverse HVS disorders and their perceptual consequences.

Purpose of the Study:

• To present a computational framework for simulating the effects of HVS disorders on perceived images.
• To leverage a modified PerceptNet model and deep learning optimization (JAX/Flax) for disease-specific simulations.
• To demonstrate the framework's utility in modeling color blindness and TBI-induced neuroplasticity.

Main Methods:

• Developed a computational framework integrating PerceptNet, a model of the HVS, with JAX/Flax deep learning.
• Modified PerceptNet to incorporate disease-specific changes relevant to HVS disorders.
• Optimized input images to match feature maps of the modified PerceptNet model.

Main Results:

• Successfully simulated visual perception alterations in color blindness.
• Demonstrated the potential to model perceptual changes in the primary visual cortex (V1) following TBI.
• Validated the framework's flexibility and robustness in simulating various HVS disorder impacts.

Conclusions:

• The developed framework provides a versatile tool for studying altered visual perception computationally.
• Offers potential applications in accessibility design, educational tools, and neuroscientific research.
• Facilitates understanding of disease-specific visual perception changes and aids in developing adaptive visual systems.