We present a new perspective on diagnostic image quality (DIQ) for mammography (MG) aimed at future clinical decision support applications. We posit that DIQ is based on three interdependent criteria: adequacy, visual quality, and interpretability. Adequacy is addressed by imaging protocols and technique, while visual quality and interpretability ensures that human perception will be able to detect diagnostically relevant content, and both enable cognitive interpretation at a conceptual level. We have implemented a visual interface for radiologists to enter annotated scores for the three criteria and overall DIQ. Score annotations may be used to identify relevant computable image features by text mining. The image features allow building DIQ metrics potentially useful during image acquisition and reading. We have conducted preliminary research on a set of relevant image features and present initial results on their discriminative power and emerging challenges. We also discuss next steps and strategies for future research.