Room: Track 1
Purpose: In digital radiography (DR), conditions of over- or underexposure are indicated to the technologist using the deviation index (DI). For DI to be meaningful, target exposure indices (TEI) appropriate for the exams need to be established first. The goal of this study was to establish TEIs for various clinical exams and to identify practice variation when TEIs cannot be established.
Methods: Approximately six months after the installation of three DR systems (Discovery XR656 HD, GE Healthcare), available exposure index data were exported, providing 20,720 data points. For body parts with at least 200 data points, the feasibility of establishing TEIs was tested by applying the flag criteria reported by Scilla et al. (AAPM 2019). A reference TEI of 250 was chosen for all exams, from which reference DI (DIref) and its standard deviation (SD), mean, skewness and kurtosis were calculated. If a body part met the flag criteria (SD(DIref)<2.5, |skew(DIref)|=1.5, and kurtosis(DIref)>0), one TEI was deemed sufficient and was derived from the mean DIref. Otherwise, a separate analysis was performed for each anatomical view.
Results: Eight body parts met the flag criteria, allowing a single TEI to be assigned (e.g., TEI(abdomen)=264 and TEI(chest)=195). Histogram analysis of DIref for chest, however, revealed more than one Gaussian distributions, necessitating view-level analysis (e.g., TEI(chest-PA)=146 and TEI(chest-LAT)=234). View-level analysis was also performed for other body parts. This identified 11 anatomical views for which TEI values could be established for the individual views (e.g., TEI(CSpine-AP)=235 and TEI(CSpine-LAO)=140). The rest of the views were analysed further based on exposure control mode, which showed in most cases that manual mode exhibited more DI variability compared to automatic exposure control.
Conclusion: The flag criteria methodology is a practical tool for establishing TEIs for clinical exams and for identifying practice variation to aid in continuous quality improvement.