This document describes a pipeline for identifying neurons from cell-based drug screening images using CellProfiler. A training set was created by labeling healthy neurons as positive examples and macrophages as negative examples. A CellProfiler rule-based classifier achieved 77% accuracy. The author also built their own classifier in R using feature transformation and nonlinear algorithms like MARS, achieving 85% accuracy. The neuron detection pipeline was validated using known toxic compounds and identified protective compounds without specific pharmacological classes.