A system and method for determining a predicted window level transformation for image data associated with a loading image based on user selections of window level transformations. A learning image associated with a set of learning context characteristics is displayed to the user in order to elicit selection of a first selected window transformation. Learned rules are established based on the learning context characteristics and the first selected window transformation. When there is a request to display a loading image associated with a set of loading context characteristics, the learned rules are evaluated based on the loading context characteristics to determine a set of inferred window level transformations. The predicted window level transformation is then determined by selecting the most desirable inferred window level transformation.