Neuronal Cell-Type Classification

Using Multimodal Features from the Allen Brain Atlas

Project Overview

This research develops machine learning models to classify neuronal cell types by combining morphological (shape) and electrophysiological (electrical activity) features. The data comes from the Allen Brain Atlas, which provides detailed measurements of neurons from both human and mouse brains.

Understanding different neuron types is crucial for studying how the brain processes information and what goes wrong in neurodegenerative disorders like Alzheimer's and Parkinson's disease. Accurate classification methods could help researchers better understand these conditions.

What I Built

• Machine Learning Pipeline: Developed and compared decision trees, random forests, and LSTM neural networks to classify neuronal cell types using multimodal features from the Allen Brain Atlas database
• Cross-Species Data Integration: Engineered a data preprocessing pipeline to harmonize human and mouse neuronal datasets, including structural layer annotations for comparative analysis
• Evaluation Framework: Implemented k-fold cross-validation and confusion matrix analysis to assess classification accuracy and identify which neuron categories are most distinguishable

Technologies Used

Python, scikit-learn, PyTorch, Allen SDK, Pandas, NumPy, Matplotlib