We are developing a first-principles task-based approach to the optimal design and evaluation of ultrasonic imaging systems. Examining five clinical features related to breast lesion diagnosis, we quantified information flow at several stages in the image formation process. We found that the diagnostic performance of a given system configuration will vary with the patient feature, sometimes significantly. Our analysis expresses diagnostic performance of an imaging system for a specific clinical task as a function of patient properties that are separable from instrument properties. Hence it is possible to show how image quality metrics, like spatial and contrast resolution, combine with patient features to determine feature discriminability. In this paper, we describe an information theoretic approach to diagnostic performance evaluation that has given us a new quantity, the acquisition information spectrum (AIS). Like NEQ in radiography, AIS in sonography provides a foundation for medical ultrasonic imaging system design.