Evaluation of subjective assessment of liver function from radionuclide images.

The appearance of non-hepatic uptake in 99Tcm-colloid images is an important indicator of poor liver function. We undertook this study to evaluate the ability of experienced observers to assess liver function from radionuclide images. A simple model was used to simulate the changing distribution of 99Tcm-colloid in the liver, spleen and bone marrow as the liver function and mass were varied. Suitable images of these three organs were isolated from real studies and used as "templates" to create realistic simulated 99Tcm-colloid images, with the count densities of liver, spleen and bone marrow determined by the model. These images were presented in random order to several observers, who were asked to assess liver function. Comparison of the estimated with the true function shows that visual assessment is systematically inaccurate, with underestimation of poor function and overestimation of good function. It is also imprecise, especially for livers with only midly impaired function. Observers were also affected by changing liver mass and were more likely to judge small livers to be abnormal. We suggest that objective assessment using a quantitative technique should be used as an adjunct to visual inspection for the evaluation of liver function.