Biology subjects

Mohammadi, Z.

Publications and source records attributed to Mohammadi, Z..

3 recordsLinked to original sources

bioRxiv2026

Strategy-dependent causal relationships between large-scale cortical activity and perceptual decisions

Animals can adopt different internally generated strategies to solve decision-making tasks, but whether cortical regions distinctly contribute to decisions under different strategies remains unknown. To address this problem, we analyzed three datasets comprising either cortical widefield imaging, or optogenetic inactivation of dozens of cortical locations in 54 mice, all performing the same sensory-evidence-accumulation task. Using Generalized Linear Model-Hidden Markov Models (GLM-HMMs), we found that the mice adopted one of three underlying strategies: an engaged, stimulus-driven state, and two disengaged states characterized by large side biases. The specific effects of inactivation of cortical regions depended on behavioral strategy: during engaged states, inactivation tended to increase side bias, while the opposite was true in biased states. For engaged states, these effects were most prominent when frontal regions were inactivated. Moreover, neural decoding analyses from cortex-wide activity revealed that frontal regions preferentially encode the animals' choices during engaged states relative to biased states. Our results suggest that distinct patterns of large-scale cortical dynamics underlie decisions under different task strategies and identify the frontal cortex as an important substrate of engaged-strategy control.

neuroscience↗

bioRxiv2024

Identifying the factors governing internal state switches during nonstationary sensory decision-making

Recent work has revealed that mice do not rely on a stable strategy during perceptual decision-making, but switch between multiple strategies within a single session [1, 2]. However, this switching behavior has not yet been characterized in non-stationary environments, and the factors that govern switching remain unknown. Here we address these questions using an internal state model with input-driven transitions. Our approach relies on a hidden Markov model (HMM) with two sets of per-state generalized linear models (GLMs): a set of Bernoulli GLMs for modeling the animals state- and stimulus-dependent choice on each trial, and a multinomial GLM for modeling input-dependent transitions between states. We used this model to analyze a dataset from the International Brain Laboratory (IBL), in which mice performed a binary decision-making task with non-stationary stimulus statistics. We found that mouse behavior in this task was accurately described by a four-state model. This model contained two "engaged" states, in which performance was good despite slight left and right biases, and two "disengaged" states, where performance was low and exhibited with larger left and right biases, respectively. Our analyses revealed that mice preferentially used left-bias strategies during left-bias stimulus blocks, and right-bias strategies during right-bias stimulus blocks, meaning that they could achieve reasonably high performance even in disengaged states simply by biasing choice toward the side with greater prior probability. Our model showed that past choices and past stimuli predicted transitions between left- and right-bias states, while past rewards predicted transitions between engaged and disengaged states. In particular, greater past reward predicted transition to disengaged states, suggesting that disengagement may be associated with satiety.

neuroscience↗

bioRxiv2022

Multichannel neural spike sorting with spike reduction and positional feature

Sorting neural voltages measured from a multichannel neural probe to extract the single unit activities of neuronal firing, especially in real-time, remains a significant technical challenge, largely due to the large amount of acquired data and the technical difficulties involved in processing and classifying these neural spikes promptly. Most neural spike sorting algorithms focus on sorting neural spikes post hoc for high sorting accuracy, and reducing the processing time generally is not the chief concern. Here we report on two signal processing modifications to our previously developed single-channel real-time spike sorting (Enhanced Growing Neural Gas) algorithm, which is largely based on graph network. Duplicated neural spikes were eliminated and represented by the neural spike with the strongest signal profile, significantly reducing the amount of neural data to be processed. In addition, the channel from which the representing neural spike was recorded was used as an additional feature to differentiate between neural spikes recorded from different neurons having similar temporal features. With these two modifications, the Graph nEtwork Multichannel (GEMsort) neural spike sorting algorithm can rapidly sort neural spikes without requiring significant computer processing power and system memory storage. The parallel processing architecture of GEMsort is particularly suitable for digital hardware implementation to improve processing speed and recording channel scalability. Multichannel synthetic neural spikes and actual neural recordings with Neuropixels probes were used to evaluate the sorting accuracies of the GEMsort algorithm.

neuroscience↗