The Role of Projected Anatomy, Random Noise, and Spatial Resolution on Clinical Image Quality in Digital Mammography

Due to the recent advances in digital detector technology, there is an increasing trend in the use of digital mammography (DM) as a concurrent modality with screen-film mammography (SFM). In order to fully realize its potential, the components of a DM system should be optimized with respect to clinical image quality in order for clinicians to make accurate diagnoses. The studies in this work involve the visual perception of medical images in an attempt to identify the major components affecting clinical image quality in DM, quantify their effects, and develop the necessary tools for carrying out such experiments.The impact of spatial resolution on image quality was evaluated via two human observer performance experiments in which the ability to determine the shape of microcalcifications was measured as a function of pixel size. Statistically significant improvements were measured between shape characterizations made at 60 µm and those made at larger pixel sizes (p 0.5).A method for simulating the appearance of malignant lesions (masses and microcalcifications) in DM images was developed and validated. The purpose of this was to design controlled visualization experiments in which the experimenter can control the exact type, shape and placement of the lesions.Using the aforementioned method, the effect of increased system noise (quantum noise, plus other noise sources inherent to the imaging system) on the detection of masses and microcalcifications was quantified via free-response human observer performance experiments. The figure of merit (FOM) for the detection of masses at the three noise-levels studied was not statistically different (p = 0.19). In contrast, the FOM for microcalcification detection decreased significantly with increasing system noise (p < 0.0001). These results indicated that smaller lesions such as microcalcifications are more prone to the effects of lower radiation dose than masses.The technique of lesion simulation was then extended to 3D in order to measure the extent to which the projected anatomy in DM affects image quality. Simulated tumors were added to images of the same patients acquired with both DM and a 3D technique referred to as breast tomosynthesis (BT). As measured through human observer detection experiments, tumors of the same size and shape were detected in BT images at 3.8 less projected signal intensity than in DM images, indicating that the projected anatomy limits the detection of tumors in DM and that BT may lead to earlier detection of breast cancer