%0 Conference Proceedings %T Unsupervised clustering of DVT Ultrasound Images using High Order Statistics %+ Lab-STICC_ENSTAB_CACS_COM %A Berthomier, Thibaut %A Mansour, Ali %A Bressollettey, L. %A Mottier, D. %A Le Roy, Frédéric %A Hermenaulty, B. %A Hoffmanny, C. %A Frechiery, L. %< avec comité de lecture %B 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) %C Madrid, Spain %I IEEE %P 2495-2501 %8 2018-12-03 %D 2018 %R 10.1109/BIBM.2018.8621187 %K Coagulation %K Veins %K Clustering algorithms %K Feature extraction %K Ultrasonography %K Elastography %Z Engineering Sciences [physics]/Signal and Image processingConference papers %X Naturally, a thrombus (or a blood clot) is developed by our body to prevent bleeding. However, various factors can slow or change the blood flow and create a thrombus in a deep vein (more often in the legs). This inappropriate situation is called Deep Venous Thrombosis (DVT) and it may permanently damage the blood vessels. Moreover, this disease (DVT) becomes deadly when a broken fragment of the thrombus reaches the lung and generates a Pulmonary Embolism. Using ultrasound images, our project aims to analyse the thrombus structure and extract information leading to its triggering factors: Immobilization, cancer, surgery, genetic variations, pregnancy, age, etc. In previous studies, we developed an approach based on wavelet transform to analyse the ultrasound images: Ultrasonography and elastography. In this manuscript, we extract new features based on High Order Statistics. These statistics provide hidden relevant information in the images to characterize the thrombus structure. %G English %L hal-02090186 %U https://ensta-bretagne.hal.science/hal-02090186 %~ UNIV-BREST %~ INSTITUT-TELECOM %~ ENSTA-BRETAGNE %~ CNRS %~ UNIV-UBS %~ ENSTA-BRETAGNE-STIC %~ ENIB %~ LAB-STICC %~ INSTITUTS-TELECOM