Optical projection tomography enables 3-D imaging of colorectal polyps at resolutions of 5-10 µm. This paper investigates the ability of image analysis based on 3-D texture features to discriminate diagnostic levels of dysplastic change from such images, specifically, low-grade dysplasia, high-grade dysplasia and invasive cancer. We build a patch-based recognition system and evaluate both multi-class classification and ordinal regression formulations on a 90 polyp dataset. 3-D texture representations computed with a hand-crafted feature extractor, random projection, and unsupervised image filter learning are compared using a bag-of-words framework. We measure performance in terms of error rates, F-measures, and ROC surfaces. Results demonstrate that randomly projected features are effective. Discrimination was improved by carefully manipulating various important aspects of the system, including class balancing, output calibration and approximation of non-linear kernels.