Extensive research has shown that practice yields highly specific perceptual learning of simple visual properties such as orientation and contrast. Does this same learning characterize more complex perceptual skills? Here we investigated perceptual learning of complex medical images. Novices underwent training over four sessions to discriminate which of two chest radiographs contained a tumor and to indicate the location of the tumor. In training, one group received six repetitions of 30 normal/abnormal images, the other three repetitions of 60 normal/abnormal images. Groups were then tested on trained and novel images. To assess the nature of perceptual learning, test items were presented in three formatsthe full image, the cutout of the tumor, or the background only. Performance improved across training sessions, and notably, the improvement transferred to the classification of novel images. Training with more repetitions on fewer images yielded comparable transfer to training with fewer repetitions on more images. Little transfer to novel images occurred when tested with just the cutout of the cancer region or just the background, but a larger cutout that included both the cancer region and some surrounding regions yielded good transfer. Perceptual learning contributes to the acquisition of expertise in cancer image perception. Significance We explored how people learned to detect tumors on cancer images. Novices classified chest radiographs over four sessions. Unlike perceptual learning of simple visual features, learning of complex features in chest radiographs supported classification of novel images. Transfer to novel images depended on the presentation of both the tumor and some of its surrounding regions; little transfer was observed when tested with just the cutout of the tumor or just the background. These results clarify the nature of perceptual learning of complex radiological images and may provide conceptual underpinning for future innovative technologies that enhance cancer image perception.