This document summarizes a study using genetic programming to classify raw chest x-ray images as either tuberculosis positive or negative, without requiring pre-processing or feature extraction steps. A three-tiered genetic programming approach was used to evolve image classification programs directly from raw pixel data. When tested on a public dataset, the genetic programming classifiers achieved classification speeds significantly faster than traditional techniques, while maintaining high accuracy levels compared to training. Visualizations showed the genetic programming classifiers focused on relevant areas of the x-ray images for classification.