“…As noted earlier, radiologists can rapidly extract the global image statistics to render an initial decision on whether a mammogram may be cancerous (Evans et al, 2016;Evans, Georgian-Smith, Tambouret, Birdwell, & Wolfe, 2013;Nodine et al, 1999). Additionally, presenting mammograms in an inverted orientation impairs radiologists' performance, suggesting that breast cancer detection is holistic (Chin, Evans, Wolfe, Bowen, & Tanaka, 2018). If perceptual learning of chest radiographs also involves the learning of global image statistics, then performance should decline when just the tumor cutout or just the background is presented.…”