Reproducibility of an imaging test: evaluation of dual liver scintigraphy by a long-term procedure

A seven-class classification system for liver scintigraphy was developed to find or exclude metastases in a clinically defined population with gastrointestinal tumours. Scintigraphy was performed twice in fifty patients a few days apart. During a period of three years each scintigraphy was classified thrice as a consensus between two experienced observers at six occasions placed on three separate days. No substantial change in classification was seen whether the same scintigraphy was interpreted twice in a day or three years apart, although considerable uncertainty was seen in interpretation of near-normal scintigraphies. In dual scintigraphies, a significant lower reproducibility was noted--directly visualized in 10 percent of the patients--indicating that a recorded difference does not necessarily reflect a change in patient situation. We think the applied procedure may be valuable for evaluation of any imaging system for clinical use.