[en] A method of identifying the dose per image when quantum mottle no longer dominates the image statistics is presented as a first step towards quantitative optimization in native and subtracted digital fluorography. The method is based on measurements of threshold contrast over a range of receptor doses and the application of a simple model of the threshold contrast detection task to estimate the magnitude of system noise sources. The point at which system and quantum noise sources are equal in magnitude is proposed as the practical upper limit for dose per image. The method is applied to a typical digital fluorography system and the results are placed into the context of the range of dose per image values found from a regional survey of digital fluorography units. While there is broad agreement between the dose per image values in the survey with values predicted from the experimental method, the considerable spread in survey doses suggests there are instances where the use of a high dose per image is unjustified. (author)