Fig. 2
From: Convolutional neural network -based phantom image scoring for mammography quality control
Unmodified (left column) test set images and modified train set images (right column) acquired with Siemens system. The ACR phantom consist of six fibres, five microcalcification groups, and five masses. A, B, and C show images acquired at 0.18 mGy, 0.31 mGy, and 4.00 mGy mean glandular doses, respectively. D and E represent degraded images with 60% and 45% noise mixing levels. F represents an average image of five ordinary QC control images. The visibility scores given for A and D were following: 1 point for fibres, 3 points for microcalcifications, and 1 point for masses (A) and 2.5 points for masses (D). The visibility scores given for B and E were following: 3 points for fibres, 4 points for microcalcifications, and 3.5 points for masses. The visibility scores for C and F were 6 points for fibres, 5 points for microcalcifications, and 5 points for masses. The visibility scores were taken from a single reviewer (VMS) scorings