[BVC (bias to variance characteristic) analysis]

The average and variance of diagnostic rating scores obtained when many physicians read one image represents the bias to variance characteristics (BVC) in image diagnosis and are parameters reflecting the quality of image that are the objects, and the physicians' characteristics in image diagnosis. They were experimentally confirmed. Combination of the results of BVC analysis with ROC analysis was shown to clarify the structure and origin of the ROC curve and to improve the interpretation of the ROC curve that was the final results of evaluation. Furthermore, it was suggested the if BVC was used as complementary data in a case study, there was a possibility of it being useful for clarify the clinical, technical and physical factors affecting the results of image diagnosis.