Methods for Automated Neuron Image Analysis

Knowledge of neuronal cell morphology is essential for performing specialized analyses in the endeavor to understand neuron behavior and unravel the underlying principles of brain function. Neurons can be captured with a high level of detail using modern microscopes, but many neuroscientific studies require a more explicit and accessible representation than offered by the resulting images, underscoring the need for digital reconstruction of neuronal morphology from the images into a tree-like graph structure.
This thesis proposes new computational methods for automated detection and reconstruction of neurons from fluorescence microscopy images. Specifically, the successive chapters describe and evaluate original solutions to problems such as the detection of landmarks (critical points) of the neuronal tree, complete tracing and reconstruction of the tree, and the detection of regions containing neurons in high-content screens.