Evaluation of Automatic Exposure Control Performance in Full-Field Digital Mammography Systems using Contrast-Detail Analysis Open Access
Downloadable ContentDownload PDF
Full Field Digital Mammography (FFDM) is increasingly replacing screen-film systems for screening and diagnosis of breast abnormalities. All FFDM systems are equipped with the Automatic Exposure Control (AEC) feature for image acquisitions. Based on the patient breast thickness, density and tissue compositions, AEC mode is meant for automatically selecting exposure technique factors to optimize dose and image quality during patient imaging. It is therefore very crucial that AEC performance is properly adjusted and optimized to different sized breasts.In this research work, I studied the AEC performance of three widely used FFDM systems. I used image quality metrics such as contrast to noise (CNR) measurements and low contrast detectability to assess image quality at different exposure levels and at different equivalent patient breast thicknesses and characterized AEC performance of the machines using the CDMAM (Contrast Detail Mammography) and QUART phantoms.Contrast-Detail (C-D) curves were generated for each AEC mode available in the FFDM systems under study for phantoms with equivalent X-Ray attenuation properties as 3.2 cm, 6 cm and 7.5 cm thick breasts. The CDMAM phantom was used to generate C-D curves that may be compared with ideal curves generated using a metric referred to as the k-factor which is the product of the thickness and the diameter of the smallest correctly identified disks in the CDMAM phantom. Previous observer studies have indicated that k-factor values of 60 to 80 µm2 are particularly useful in demonstrating the threshold for object detectability for detectors used in digital mammography systems. Thus, my results were compared to ideal C-D curves corresponding to these k-factors. The QUART Phantom was used to calculate CNR values at different phantom thicknesses. The results of the C-D analysis and CNR measurements were used to determine limiting CNR values intended to provide a threshold for proper image quality assessment.The results of the Contrast-Detail analysis show that for higher phantom thicknesses, low contrast signal detectability gets worse. This agrees with the results obtained with the QUART phantom, where CNR decreases below determined limiting CNR values.My results show that signal detectability gets worse on average by 27 % in the 5 cm thick phantom and by 56 % in the 6 cm thick phantom when compared to signal detectability in the 3 cm thick phantom when using the AEC in two of the three evaluated FFDM systems. This indicates that when using the AEC in these systems, low contrast breast abnormalities may not be visualized in thick breasts due to a deterioration of image quality at the exposure level selected. This raises the concern that the AEC performance in these two FFDM systems is only optimized for standard breast thicknesses and low contrast detectability is compromised at higher thicknesses because of inadequate exposure. Therefore, I developed technique charts for these systems containing optimized exposure parameters for all three phantom thicknesses.