Detectability of T1a lung cancer on digital chest radiographs: An observer-performance comparison among 2-megapixel general-purpose, 2-megapixel medical-purpose, and 3-megapixel medical-purpose liquid-crystal display (LCD) monitors

Background: There has been no comparison of detectability of small lung cancer between general and medical LCD monitors or no comparison of detectability of small lung cancer between solid and part-solid nodules. Purpose: To compare the detectabilities of T1a lung cancer on chest radiographs on three LCD monitor types: 2-megapixel (MP) for general purpose (General), 2-MP for medical purpose (Medical), and 3-MP-Medical. Material and Methods: Radiographs from forty patients with T1aN0M0 primary lung cancer (27 solid nodules, 13 part-solid nodules) and 60 patients with no abnormalities on both chest X-ray and computed tomography (CT) were consecutively collected. Five readers assessed 100 cases for each monitor. The observations were analyzed using receiver operating characteristic (ROC) analysis. A jackknife method was used for statistical analysis. A P value of <0.05 was considered significant. Results: The average AUC for all T1a lung cancer nodule detection using the 2-MP-General, 2-MP-Medical, and 3-MP-Medical LCD monitors were 0.86, 0.89, and 0.89, respectively; there were no significant differences among them. The average AUC for part-solid nodule detection using a 2-MP-General, 2-MP-Medical, and 3-MP-Medical LCD monitors were 0.77, 0.86, and 0.89, respectively. There were significant differences between the 2-MP-General and 2-MP-Medical LCD monitors (P=0.043) and between the 2-MP-General and 3-MP-Medical LCD monitors (P=0.027). There was no significant difference between the 2-MP-Medical and 3-MP-Medical LCD monitors. The average AUC for solid nodule detection using a 2-MP-General, 2-MP-Medical, and 3-MP-Medical LCD monitors were 0.90, 0.90, and 0.88, respectively; there were no significant differences among them. The mean AUC values for all and part-solid nodules of the low-experienced readers were significantly lower than those of the high-experienced readers with the 2 M-GP color LCD monitor (P<0.05). Conclusion: Detectability of part-solid nodules using a general-purpose LCD monitor was significantly lower than those using medical-purpose LCD monitors.