The detection of image features is an essential component of medical image processing, and has wide-ranging applications including adaptive filtering, segmentation, and registration. In this paper, we present an information-theoretic approach to feature detection in ultrasound images. Ultrasound images are corrupted by speckle noise, which is a disruptive random pattern that obscures the features of interest. Using theoretical probability density functions of the speckle intensity distributions, we derive analytic expressions that measure the distance between distributions taken from different regions in an ultrasound image and use these distances to detect features. We compare the technique to classic gradient-based feature detection methods.