A new energy detector of micro-emboli using a time-varying threshold

International audienceDetection of small emboli, precursors of Cerebrovascular Accidents, is a worldwide concern since CVAs represent the second cause of mortality in the world. Computerized analysis of Transcranial Doppler signals can aid early detection of circulating emboli and micro-emboli. Commercially used systems of automatic emboli detection rely on standard short time Fourier transform techniques in which detection is based on constant thresholds. These standard algorithms do not offer robust detections and are incapable of detecting the smallest micro-emboli. To enhance this detection, we propose in this study optimized techniques based on novel methods of threshold application. By implementing our new time-varying threshold of detection, we were able to decrease the probability of non-detection and the probability of false alarm by around half the values obtained by standard techniques. Moreover, our new techniques were clearly efficient in exploiting the transient-like embolic signals and hence make detection of micro-emboli easier and more evident. This was proved by enhancing important parameters of which are the embolus to blood ratio and the peak to threshold ratio. Applied on our set of recorded signals, the new detectors allowed obtaining embolus to blood ratios twice greater than the embolus to blood ratios achieved by standard techniques and a sufficient increase in peak to threshold ratios