Measurement of cerebral blood flow velocity (CBFV) by transcranial Doppler has become a common clinical tool to evaluate the dynamic changes of cerebral blood flow. CBFV pulse waveform morphology can be also used for the non-invasive estimation of intracranial pressure. Reliable tracking of CBFV morphology requires accurate beat-to-beat delineation. However, CBFV is inherently contaminated with various types of noise and artifacts that obscure its pulse delineation. Adaptive thresholding (ADT) is a promising method originally developed for the onset detection of photoplethysmogram signal. The current work studies the application of ADT method for the CBFV pulse onset detection. A large annotated dataset of 92,794 CBFV pulses collected from 108 subarachnoid hemorrhage patients, admitted to UCLA Medical Center, were employed in the study. The ADT method was applied to the dataset and its performance in terms of true positive rate (TPR) and positive predictivity value (PPV) was evaluated. We also proposed some modifications to the original algorithm to enhance its efficacy for the CBFV onset detection. The results showed that while the original ADT achieved TPR of 75.9% and PPV of 82.7% in detection of CBFV pulse onsets, our proposed modifications significantly enhanced the performance levels to TPR of 93.1% and PPV of 93.3%, respectively. We conclude that the modified adaptive thresholding algorithm could be reliably used for the pulse onset detection of the CBFV signal.