Computer based pattern recognition of carotid artery Doppler signals for disease classification: prospective...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Cardiovascular Diagnostics

Background:

• Internal carotid artery (ICA) stenosis is a significant risk factor for stroke.
• Accurate classification of stenosis severity is crucial for patient management.
• Current diagnostic methods like angiography can be invasive.

Purpose of the Study:

• To develop and validate a computer-based pattern recognition method for classifying internal carotid artery stenosis.
• To assess the accuracy of this non-invasive method against conventional angiography.

Main Methods:

• Utilized a combined B-mode/pulsed Doppler unit to obtain spectral waveforms.
• Employed ECG-R wave for synchronization and averaged Doppler spectra from 20 heart cycles.
• Applied a stepwise selection algorithm to extract features and partition disease severity into categories: Normal, 1-20%, 21-50%, and 51-99% diameter reduction.

Main Results:

• The computer system achieved 82% overall agreement with angiography in classifying 170 vessels.
• High agreement was observed across all categories: 93% for normal, 81.5% for 1-20%, 78% for 21-50%, and 82% for 51-99% stenosis.
• Discrepancies of more than one category occurred in only one case, with a chance-corrected agreement (Kappa) of 0.769.

Conclusions:

• The developed computer-based pattern recognition method provides a reliable and accurate non-invasive approach for classifying internal carotid artery stenosis.
• This system demonstrates significant potential for routine clinical use and epidemiological studies.
• Future work includes refining classification, developing dedicated hardware, and prospective validation.