اميرسالار نطاق

today, with the industrialization of the developed communities changing the habits of human life and adding various high-risk behaviors to human health, congestion of arteries, and subsequently, heart and heart disease has become one of the most challenging issues related to the health of the society, to the extent that the blockage of the heart and constriction of coronary arteries is one of the most common and common causes of mortality worldwide. diagnostic methods are divided into two categories: aggressive and non-invasive. Non-invasive diagnostic methods are more popular because they are not vulnerable to the patient. different imaging techniques are one of the most common non-invasive methods. therefore, various imaging techniques have been used to diagnose and treat the disease. angiography is a common imaging technique that presents the medical duct of the vessel but is unable to show plaque shape, plaque components, and plaque contents. The intravascular ultrasound is a well-known and effective method of imaging. in this method by sending sound waves at a 20-40 Mhz frequency of through the catheter to the lining of the veins and back of the waves from these layers, a continuous image can be created which not only provides useful information about the internal and external walls of the vessel but also provides the shape, size, and type of the blocker. this method is suitable for diagnosis and selection of treatment methods. using deep neural networks in most image processing processes has been associated with good results and in most of the studies in various fields related to machine vision the best results. in this thesis, a novel method for automatic segmentation of IVUS cells in IVUS images has been proposed using deep neural networks. The proposed method in this thesis has been eva‎luated on two data sets for further eva‎luation, first, the intra-array ultrasound database of Boston, which is one of the most well-known databases in this field and belongs to Boston University, and the second is the Valcano database. Finally, achieving an average accuracy of 93.5% for MA and92.2% for Lumen on the Boston intravascular ultrasound database, and an average accuracy of 94.3% for MA and 95.3% for Lumen on the Volcano database is a testimony of good performance. It is a recommended method.

تنگ شدن عروق كرونري , اولتراسوند درون عروقي , شبكه عصبي عميق , مرزبندي خودكار عروق.

Coronary artery narrowing , intravascular ultrasound , deep neural network , Automatic demarcation of vessels

Amirsalar Nattagh

Dr.Ahmad Ayattolahi