Emboli detection in middle cerebral artery using continuous wavelet transform

Gonçalves, Ivo Manuel Belchior

http://hdl.handle.net/10400.1/3114

Use this identifier to reference this record.

Name:	Description:	Size:	Format:
EMBOLI_DETECTION_IN_MIDDLE_.PDF		909.88 KB	Adobe PDF	Download

Send Feedback

Authors

Gonçalves, Ivo Manuel Belchior

Advisor(s)

Leiria, Isabel

Moura, Maria Margarida Madeira

Abstract(s)

Doppler ultrasound is a common non-invasive technique for data acquisition for assistance in clinical analysis and diagnosis. Amongst the clinical applications to blood flow is the detection and characterization of emboli. In this context it was made a comparative study between spectral estimation methods in detection of embolic events in middle cerebral artery blood flow signal. The comparison is made between the Short-Time Fourier Transform and the Continuous Wavelet Transform, whereas the former is the classical and widely discussed method and the latter appears as promising alternative given the known characteristics, such as for non-stationary signal processing. To the signals studied, generated using a simulator, it were added simulated embolic events with known characteristics, like duration, localization and power. The classical method was configured in order to obtain reference results for comparison, whereas the alternative method studied was used with multiple configurations, enabling a richer study on the parameters variation. The analysis to the results obtained for each method was made in terms of false negatives, false positives and sensitivity. The results show that CWT, presented as alternative methods, allows a better embolic events detection rate than STFT, the classical method. All CWT config- urations studied demonstrate better results than STFT. Differences in the results for CWT allow to draw conclusions about the best configuration, among the ones studied, for results optimization, being that the configuration based on Mexican Hat is the one which presents overall better results.