The reliability (precision) and validity (accuracy) measurements of lung nodule size were part of my investigations of lung cancer screening and detection lung nodules using CT.   The rational was that nodule growth rate is a measure of malignancy and nodule size is used for medical decisions.   Reliability and validity were evaluated using a phantom that could be placed on the patient's chest.  

The first description of a lung nodule phantom was the 2000 Medical Physics World Congress  in Chicago.  The 2001 Proceedings paper from SPIE Medical Imaging  was for a poster. The proceedings paper describes in detail the final design of the phantom. (Ball and cubes - 2mm to 8mm)   

A comparison of reliability (standard deviation of measurement) of computer measurements of phantom nodules and human evaluations was made on the same images.  Computer measurement were  10 times more reliable than human measurements. In addition, the reliability of human measurements on nodule-like features in patient images was determined to be similar to phantom nodule reliability.   The comparison was presented at the RSNA 2002 Reliability.