Mircea Teodorescu

PULSE, a foundation model trained on neural recordings, reveals how disease variants affect network development and how circuits encode information in brain organoids.

The paper introduces MYCEL, a graph neural network that improves decoding of brain anatomy from neuronal neighborhoods by learning local pooling, outperforming fixed aggregation on multiple datasets.

HIPPIE is a deep learning model that classifies neurons from electrophysiological recordings by combining self-supervised pretraining and supervised fine-tuning, showing better performance than unsupervised methods.

The paper presents a cloud-based pipeline for analyzing and visualizing neuronal electrophysiology data from HD-MEAs, demonstrating its utility in simplifying data analysis and understanding neural activity.

This paper develops a cloud-connected stem cell platform to enhance scientific identity in underrepresented students, showing improved outcomes in competence and research readiness.

This paper develops a 3D-printed, servo-actuated microvalve system for automated organoid culture, validated in mouse and human models with comparable viability and gene expression to manual methods.

This paper introduces a microfluidic platform called directoids to study directional axonal growth between neural organoids, showing robust directional bias in axon outgrowth.

SpikeLab is a text-to-analysis framework for neural spike data that ensures correct and reproducible results using agentic systems, outperforming other models in a controlled benchmark.

This study develops a microfluidic platform to induce spatially segregated ventral and dorsal domains in mouse forebrain organoids without fusion, enabling study of tissue regionalization.

This study uses mouse forebrain organoids to show how variations in cellular composition influence the self-organization of neural circuits, revealing intrinsic principles of forebrain circuit assembly.

This paper introduces an automated in-incubator microscope for long-term live cell imaging, enabling high-frequency, continuous data collection with improved thermal stability and operational reliability.

This study demonstrates goal-directed learning in cortical organoids using a closed-loop electrophysiology framework and artificial reinforcement learning.

This paper introduces vascularized cortical assembloids (vCAs) that model neurovascular co-development and show improved cellular specialization and connectivity compared to standard cortical organoids.

This study compares fluoxetine delivery via a bioelectronic device to topical bolus dosing in swine wounds, showing higher tissue concentrations and safety with the device.

This study shows that structured firing sequences emerge in human and murine brain organoids and neonatal brain slices, suggesting innate rather than experience-dependent neural coding.

This paper develops a wireless bioelectronic actuator for wound healing in pigs, showing reduced M1/M2 macrophage ratio and increased re-epithelialization in treated wounds.

This paper introduces a sealed recirculatory system for organoid culture that eliminates the need for CO2 incubators and enables continuous live imaging and metabolic viability.

This paper develops a deep reinforcement learning framework for real-time wound healing treatment optimization, showing 17.71% acceleration in in silico experiments and 40% in vivo.

This paper introduces an automated brain organoid platform with IoT integration for continuous monitoring and control, showing increased neural activity and stable firing rates during automated media exchange.

This paper develops a remote-controlled wireless bioelectronic system for delivering fluoxetine to promote wound healing in pigs, showing improved re-epithelialization and reduced inflammation.

This study shows that the identity of MGE-derived cortical interneurons can be dynamically influenced by their environment, with human 3D models promoting Pvalb expression over Sst.

The paper shows that slower axon growth in humans, driven by reduced calcium influx, leads to longer axon tracts through prolonged growth duration, with calcium dynamics regulating developmental timing.

The paper presents a wearable imaging system for automated wound monitoring in pigs, validated to capture and analyze wound progression using machine learning.

The paper uses high-density electrophysiology to study epileptiform activity in human hippocampal slices from epilepsy patients, revealing unexpected network oscillations in dentate gyrus slices.

SIMS is a deep learning tool for single-cell RNA sequencing classification that outperforms existing methods in accuracy and versatility across complex tissues and developmental stages.

The paper demonstrates the use of a sliding mode controller in a bioelectronic device to deliver a pre-determined dosage of fluoxetine with precision over time.

The paper introduces a new transfer function called Refractory SoftPlus to improve mean-field models of neuronal networks, enabling accurate predictions of network responses and bifurcation analysis without restrictive assumptions.

The paper introduces Piphys, an open-source platform for neural recording that enables remote control and scalability through cloud-based software and low-cost hardware.