[Automated detection of microemboli in transcranial Doppler signals]

The detection of micro-emboli (ME) in the intracranial brain arteries by means of transcranial Doppler sonography depends on signal quality and on the definition of the detection levels. The qualitative analysis by ear should be replaced by quantitative measurement techniques to provide results that are cross-comparable to the ones of other studies. Algorithms for ME detection have been developed and implemented on a Doppler signal analyzer. Three quantitative criteria were established with test data. Tape recordings of 44 middle cerebral arteries from an ongoing study including patients before and after implantation of prosthetic heart valves were examined. The data were first corrected for frequency response, phase and amplitude and examined by two independent investigators and by the instrument. Within the 44 measurements both examinations by ear found the same 26 samples to contain zero audible ME signals. The instrument detected in 21 cases zero ME signals, in 4 cases 1 signal and in 1 case 4 signals. For 18 tapes the ear-method provided counts of 1 to 160 ME's. In 12 cases the two investigators got the same result, in the remaining 6 cases the higher of both figures was selected for comparison to the automated count. The counting by the instrument was exactly the same for 7 cases, in 7 cases the instrument counted more, in 4 cases it detected fewer ME signals compared to the reference. The method and the proposed detection criteria provided more false positive than false negative results. This appeared to be due to artifact detection and on higher resolution in time leading to better separation of double events compared to the ear method. Assuming an adequate quality of the quadrature signals, the method may lead to automatic screening for ME signals.