COMPUTER - AIDED PROCESSING AND ANALYSIS OF ULTRASOUND IMAGE SEQUENCES FOR THE DETECTION AND CHARACTERIZATION OF THYROID NODULESPERSON IN CHARGE: Maroulis Dimitris
FUNDED BY: GGET
START DATE: 01/04/2005
END DATE: 30/09/2008
The objective of the project is to study and develop novel methods for the processing and analysis of ultrasound image sequences, aiming on the detection and characterization of thyroid nodules. These methods will be integrated on a computer system for the support of medical diagnosis. The system shall feature high sensitivity and specificity, minimizing the false detection likelihood. Pattern recognition methodologies shall be applied, including feature extraction from the ultrasound images, feature selection and classification. The features to be investigated include: 1) texture features and 2) color features extracted from Doppler ultrasound images. Classic and modern learning machine algorithms shall be tested for the classification. Moreover, time-consuming algorithms shall be implemented on a FPGA device to achieve real time response. The system performance shall be evaluated using unbiased methods widely accepted from the scientific community.