Elevation of intracranial pressure (ICP) may occur in many diseases and therefore the ability to measure it non-invasively would be useful. Flow velocity signals from Transcranial Doppler (TCD) have been used to estimate ICP, however the relative accuracy of these methods is unclear. This study aimed to compare 4 previously described TCD-based methods with directly measured ICP in a prospective cohort of head injured patients. Non-invasive ICP (nICP) was obtained using the following methods: I) a mathematical “black-box” model based on interaction between TCD and ABP (nICP_BB); II) based on diastolic FV (nICP_FVd); III) based on critical closing pressure (nICP_CrCP) and IV) based on TCD-derived pulsatility index (nICP_PI). In time domain, for recordings including spontaneous changes in ICP greater than 7 mmHg, nICP_PI showed the best correlation with measured ICP (R=0.61). Considering every TCD recording as an independent event, nICP_BB generally showed to be the best estimator of measured ICP (R=0.39, p<0.05; 95% CI=9.94 mmHg; AUC= 0.66, p<0.05). For nICP_FVd, although it presented similar correlation coefficient to nICP_BB and marginally better AUC (0.70, p<0.05), it demonstrated a greater 95% CI for prediction of ICP (14.62 mmHg). nICP_CrCP presented a moderate correlation coefficient (R=0.35, p<0.05) and similar 95% CI to nICP_BB (9.19 mmHg), but failed to distinguish between normal and raised ICP (AUC=0.64, p>0.05). nICP_PI was not related to measured ICP using any of the above statistical indicators. We also introduced a new estimator (nICP_Av) based on the average of 3 methods (nICP_BB, nICP_FVd and nICP_CrCP), which overall presented improved statistical indicators (R=0.47, p<0.05; 95% CI=9.17 mmHg; AUC= 0.73, p<0.05). nICP_PI appeared to reflect changes in ICP in time most accurately. nICP_BB was the best estimator for ICP ‘as a number’. nICP_Av demonstrated to improve the accuracy of measured ICP estimation.