[ROC analysis for evaluating the detectability of image unsharpness due to the patient's movement: phantom study...

To evaluate the detectability of image unsharpness due to a patient's movement, a receiver operating characteristic (ROC) analysis was conducted to compare the diagnostic and preview liquid-crystal displays (LCDs). Phantom images that simulated a patient's movement were obtained by using a moving metronome and acrylic plates with a computed radiography (CR) system. A total of 104 images were classified into five groups according to the degrees of image unsharpness determined based on the metronome velocity and exposure time. In an ROC observer study (n=6), a 2-megapixel diagnostic monochrome LCD (2M-LCD) and a 1.3-megapixel general color LCD for preview (1.3M-LCD) were compared in terms of the detection of image unsharpness due to the movement. A statistical test was performed using the multi-reader multi-case (MRMC) method. In the results, the average areas under the ROC curve values for the detection of image unsharpness using the 2M-LCD and 1.3M-LCD were 0.952 and 0.850, respectively. The detection of image unsharpness using the 2M-LCD was significantly better than that using the 1.3M-LCD (p<0.05). In addition, some images with slight unsharpness were identified correctly only using the 2M-LCD. The results suggest that the low-resolution LCD (i.e., the 1.3M-LCD for preview) had a limitation in identifying image unsharpness due to the patient's movement. Slight unsharpness could be missed in primary image checks performed on a preview monitor equipped with an imaging system. Therefore, the high-resolution LCD (i.e., a 2M-LCD) is necessary when using radiography for diagnostics.