Feasibility and reliability of on-line automated microemboli detection after carotid endarterectomy. A transcranial Doppler study.

OBJECTIVES: recently, a new algorithm for transcranial Doppler (TCD) ultrasound detection of microembolic signals (MES) was developed. In the present study, we investigated its on-line performance in TCD monitoring after carotid endarterectomy (CEA) and assessed off-line its accuracy in detecting MES. MATERIALS AND METHODS: first, the feasibility of MES detection in TCD monitoring after CEA in a routine clinical setting was evaluated in 50 patients. Second, to test the reliability of the software a 2-h digital audio study tape was made and analysed by the algorithm and five human experts. The "gold standard" was defined as the agreement between human experts: a MES was considered to be present if at least three human observers agreed. RESULTS: TCD monitoring for emboli detection after CEA was well tolerated by the patients and could be performed reliably. In the study tape, the human gold standard detected 107 MES, with 93 MES having an intensity of > or =7 dB. The software detected 81 and 77 MES, respectively. Using the 7 dB intensity threshold, the software had no false positives and 16 false negatives. The kappa value between the human gold standard and the software was 0.91, the proportion of specific agreement was 0.83. CONCLUSIONS: the tested algorithm provides a reliable method for automated on-line microemboli detection after CEA. This makes monitoring of the effectiveness of antiplatelet agents in the prevention of stroke after CEA more practicable.