Although Lesion size index (LSI) has been reported to highly predict radiofrequency lesion size in vitro, its accuracy in lesion size and steam pop estimation has not been well investigated for every possible scenario. Initially, radiofrequency ablations were performed on porcine myocardial slabs at various power, CF, and time settings with blinded LSI. Subsequently, radiofrequency power at 20, 30, 40, 50, and 60W was applied at CF values of 5, 10, 20, and 30g to reach target LSIs of 4, 5, 6, and 7. Lesion size and steam pops were recorded for each ablation. Lesion size was positively correlated with LSI regardless of power settings (p<0.001). The linear correlation coefficients of lesion size and LSI decreased at higher power settings. At high power combined with high CF settings (50W/20 g), lesion depth and LSI showed an irrelevant correlation (p=0.7855). High-power ablation shortened ablation time and increased the effect of resistive heating. LSI could predict the risk of steam pops at high-power settings with the optimal threshold of 5.65 (sensitivity, 94.1%; specificity, 46.1%). The ablation depth of the heavy heart was shallower than that of the light heart under similar ablation settings. LSI could predict radiofrequency lesion size and steam pops at high power settings in vitro, while synchronous high power and high CF should be avoided. Lighter hearts require relatively lower ablation settings to create appropriate ablation depth.