Accurate and Automatic Carotid Plaque Characterization in Contrast Enhanced 2-D Ultrasound Images

The carotid plaques characterization is essential to decide about the possibility of surgical intervention (endarterectomy/stenting) on the patient. Soft and unstable plaques represent a major risk for the patient, as they are correlated with an augmented probability of brain infarction and emboli generation. Hence, the minimally-invasive characterization expecially of this type of carotid plaques is crucial in clinical practice. This paper presents an integrated system for the completely user-independent carotid plaque segmentation and characterization, based on ultrasound 2-D images. We show that using a ultrasound contrast agent, it is possible to segment also echolucent plaques with a percentage of misclassified pixels equal to 8%. After segmentation, the enhanced image is used to perform tissue characterization. We tested our system on 5 echolucent plaques and on 5 fibrous/stable plaques, showing that our system is capable of an accurate carotid wall segmentation and proper quantification of the percentages of blood, fat, calcium and fibrous tissue constituting the plaque. The system is very promising and it is being used in a neurology unit on patients already indicated for endarterectomy, with the purpose of correlating its output with histology.