Model-Free Estimation of Tuning Curves and Their Attentional Modulation, Based on Sparse and Noisy Data

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Sensory neuron responses are often modeled using tuning curves, typically fitted with bell-shaped functions to discrete stimulus values.
• Conventional model fitting struggles with irregular neuronal responses and noisy data, leading to unreliable model selection.
• Model choice significantly impacts quantitative and qualitative comparisons of neuronal activity across experimental conditions.

Purpose of the Study:

• To address the challenges of model selection in analyzing neuronal tuning curves.
• To introduce and validate a robust, model-free approach for extracting tuning features from neuronal recordings.
• To investigate attentional modulation of neuronal responses in area MT using both conventional and novel methods.

Main Methods:

• Fitting diverse mathematical models (e.g., Gaussian) to neuronal response data from area MT in rhesus monkeys.
• Developing and applying a model-free, data-driven approach to extract tuning curve features directly from response data.
• Analyzing responses to complex composite stimuli during various attentional tasks.

Main Results:

• All tested models could be well-fitted, but the best model varied significantly between neurons.
• Model-free methods provided results consistent with fit-based approaches when good fits existed.
• Model-free methods revealed attentional modulation patterns, including alterations in irregular tuning curve shapes, missed by conventional models.
• Attentional effects were found to be cell- and stimulus-specific.

Conclusions:

• Model-free approaches offer a reliable and robust alternative to conventional model fitting for analyzing neuronal tuning curves.
• These data-driven methods can extract relevant tuning information even from neurons with irregular response patterns.
• The findings highlight the cell- and stimulus-specific nature of attentional modulation in sensory cortex.