This work describes research towards computerassisted diagnostic neuropathology using support vector machines. The system processes digital photomicrographs in three phases: image preprocessing, feature extraction and classification. Compactness, fractal dimensions and co-occurrence matrices were used to generate a feature vector and then three different kernels were used for support vector machine classification. The results showed an increase in the system?s performance when a combination of features was used rather than a single feature, with best results obtained for combination of all three features. Despite limited data, the results were promising, suggesting that this approach is worth exploring further.