Real-time identification and archiving of micro-embolic Doppler signals using a knowledge-based DSP system.

Identification of micro-emboli in the cerebral circulation using transcranial Doppler ultrasound provides valuable clinical information, but, currently, embolic signal detection and analysis are significantly limited because they mainly rely on costly off-line analysis by human experts. In this study, a reliable, high-resolution, real-time automated system for the detection and archiving of embolic signals was designed and implemented using expert system theory and modern DSP technology. Preliminary tests were conducted to evaluate the functions and the performance of the system using data from ten carotid endarterectomy patients and two normal volunteers. Using the widely accepted 7 dB threshold for human reliability and a human expert, majority-decision gold standard, the real-time system reached sensitivity and specificity of 93.6% and 99.3%, respectively, which were close to the results obtained by three human experts under ideal laboratory conditions (90.1% and 99.8%, 98.4% and 99.9%, 98.9 and 99.9%). The new system has the potential to be used either as a bedside monitoring and signal acquisition device, or as a laboratory investigation tool.