Goal-Directed Learning in Cortical Organoids.

Area of Science:

• Neuroscience
• Stem Cell Biology
• Computational Neuroscience

Background:

• Advancements in electrophysiology and electrical stimulation are key to understanding brain function.
• Cortical organoids from pluripotent stem cells are valuable in vitro models for brain development.
• The impact of sustained, meaningful sensory input on in vitro neural cultures remains underexplored.

Purpose of the Study:

• To demonstrate goal-directed learning in brain organoids.
• To investigate the effects of different training signal selection methods on learning.
• To establish a framework for studying in vitro learning mechanisms.

Main Methods:

• Developed a closed-loop electrophysiology framework.
• Integrated mouse cortical organoids into a simulated 'Cartpole' task.
• Utilized high-frequency training signals and longitudinal experiments.

Main Results:

• Organoids trained with artificial reinforcement learning showed improved task performance.
• Reinforcement learning outperformed random signal selection and no training.
• Demonstrated the feasibility of evaluating learning through simulated tasks.

Conclusions:

• Goal-directed learning is achievable in brain organoids.
• Artificial reinforcement learning is an effective method for training organoids.
• This approach advances in vitro neuroscience for therapeutic and computational applications.