Marius Pachitariu

Raster photostimulation of large-scale neural populations

This paper introduces a new system for activating and recording thousands of neurons over a large field-of-view with pixel-level precision.

A critical initialization for biological neural networks

The paper shows that large-scale neural recordings in mice exhibit dynamical properties similar to those of a critically normalized random symmetric matrix, suggesting a critical initialization for biological neural networks.

Linking neural representations to behavior using generalization

The paper links neural representations to behavior by testing mice's responses to new stimuli and finding that neural discrimination correlates with behavioral performance in specific visual areas.

Live Spike Sorting of Large-scale Neural Recordings

This paper introduces a live spike sorting system that achieves performance comparable to offline sorting and enables closed-loop interventions in neural recordings.

Extracting large-scale neural activity with Suite2p

Suite2p processes large-scale neural recordings, outperforming existing methods in motion correction and cell detection.

Orofacial behaviors, not eye movements, govern neural activity in mouse visual cortex

This study shows that orofacial movements, not eye movements, primarily drive neural activity in mouse visual cortex during free viewing.

High performance sorting of motor unit action potentials with EMUsort

EMUsort improves MUAP spike sorting accuracy by addressing challenges like long delays, complex waveforms, and high overlap in motor unit recordings.

Spatial predictive coding in visual cortical neurons

The paper tests whether mismatch responses in visual cortex are due to spatial or temporal predictive coding by manipulating visual stimuli in mice.

Large-scale high-density brain-wide neural recording in nonhuman primates

The paper introduces Neuropixels 1.0-NHP, a high-density electrode array for large-scale neural recording in nonhuman primates, enabling recording from over 3000 single neurons.

Cellpose-SAM: superhuman generalization for cellular segmentation

Cellpose's flow-field output on SAM's pretrained encoder yields one segmenter that errs less than a second human annotator on out-of-distribution images (0.163 vs 0.257 error rate).

Cortical control of innate behavior from subcortical demonstration

This study shows that mice can learn to avoid a cold stimulus through cortical control, even without prior experience, suggesting a dynamic interaction between cortical and subcortical brain areas.

Cellpose3: one-click image restoration for improved cellular segmentation

Cellpose3 improves cellular segmentation in noisy, blurry, or undersampled microscopy images through one-click image restoration tools.

Rastermap: a discovery method for neural population recordings

Rastermap is a visualization method for neural data that sorts neurons along a one-dimensional manifold to reveal activity patterns in high-dimensional recordings.

Assembloid model to study loop circuits of the human nervous system

The paper develops an assembloid model of the human cortico-striatal-thalamic-cortical circuit to study its function and disease-related disruptions.

Equivariant neuronal populations enable simultaneous tuning and invariance

The paper investigates how neuronal populations encode visual scenes invariant to perspective changes, finding equivariant structures in higher-order visual areas.

Towards a simplified model of primary visual cortex

The paper simplifies artificial neural network models of primary visual cortex while maintaining predictive power, using data from mouse and monkey V1 neurons.

Benchmarking cellular segmentation methods against Cellpose

The paper benchmarks cellular segmentation methods, showing Cellpose outperforms Mediar when tested fairly and a transformer-based variant of Cellpose does not improve performance.

Distinct streams for supervised and unsupervised learning in the visual cortex

The paper investigates whether perceptual learning in mice is due to supervised or unsupervised learning by analyzing neural activity in visual cortex during task learning and unrewarded exposure.

Anatomically distributed neural representations of instincts in the hypothalamus

The paper shows that distinct, anatomically distributed neuron ensembles in the hypothalamus encode social and fear behaviors through mixed selectivity, with strong localization bias.

Facemap: a framework for modeling neural activity based on orofacial tracking

Facemap models neural activity using orofacial tracking, improving prediction accuracy and explaining more variance in visual cortex neurons.

Solving the spike sorting problem with Kilosort

Kilosort4 swaps K-means-style clustering for a bipartite modularity graph method with a merging tree, and recovers most ground-truth neurons in simulations where competing sorters lose half of them.

Cellpose 2.0: how to train your own model

Cellpose 2.0 enables training specialist segmentation models with minimal user effort and data, achieving state-of-the-art results on new image categories.

Structured random receptive fields enable informative sensory encodings

The paper models neuronal receptive fields as random samples from distributions, showing improved performance in sensory encoding tasks.

Toroidal topology of population activity in grid cells

The paper shows that grid cell activity forms a toroidal manifold, supporting continuous attractor network dynamics in the medial entorhinal cortex.

Electrode pooling: How to boost the yield of switchable silicon probes for neuronal recordings

This paper introduces electrode pooling to increase the number of recording sites on silicon probes by using a single wire for multiple sites, demonstrating its feasibility on Neuropixels 1.0 and simulating its effectiveness.

Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings

The paper introduces Neuropixels 2.0, a high-density probe for long-term brain recordings, and demonstrates its stability across months in multiple labs and species.

High precision coding in mouse visual cortex

The paper measures neural population coding precision in mouse visual cortex and finds that discrimination thresholds are much lower than behavioral thresholds, suggesting downstream decoding limitations.

A survey of spiking activity reveals a functional hierarchy of mouse corticothalamic visual areas

The paper presents a dataset and analysis of spiking activity across multiple mouse visual areas, revealing that functional connectivity mirrors anatomical hierarchy.

Cellpose: a generalist algorithm for cellular segmentation

Cellpose segments diverse 2D and 3D microscopy cell images without retraining by predicting topological gradient vector flows via a style-conditioned residual U-net and routing pixels to cell centers.

Retinal outputs depend on behavioural state

The paper shows that retinal outputs in mice are modulated by behavioural state, as observed in the superficial superior colliculus and optic tract, even in darkness.