Mohseni Sayyed Ourmazd, DDS

Utilization of a 3D Convolutional Neural Network for Automated Volumetric Analysis of Alveolar Clefts

Sayyed Ourmazd Mohseni is a researcher and intern at UT Southwestern Medical Center/Parkland Hospital. He completed his B.Sc. in Biological Sciences with an emphasis on computational biology from ASU; Graduating Summa Cum Laude and receiving the Moeur award for highest academic achievement. He continued his education at USC school of dentistry during which he was engaged with clinical research projects; Leading to several publications and presentations. Upon conferral of his DDS degree, he was inducted into the OKU honor society and continued his passion for research through a dedicated research position at UTSW, focusing on tissue engineering and cancer diagnostics. He hopes to apply his research training to topics surrounding OMFS.

Abstract:
Cleft lip and palate (CLP) are one of the most common facial defects treated by oral and maxillofacial surgeons with a reported prevalence of 1 in every 1600 babies in the United States. Accurate identification of the alveolar cleft site along with volumetric analysis provides a foundation for accurate surgical planning and improved post-operative outcomes. Recent advancements in artificial intelligence and deep learning provide a promising avenue for highly accurate surgical planning. Accurate identification of alveolar cleft landmarks and calculation of volumes remains a challenging concept due to subjectivity and variation of landmark identification by providers with experienced surgeon annotation being the current gold standard. 3D Convolutional Neural Networks (CNNs) have been previously utilized for identification of both anatomical structures and pathology in radiology with high levels of accuracy; However, their utilization remains limited in craniofacial surgical planning. This study aims to develop, train and utilize 3D CNNs for automated identification of alveolar cleft defects in an attempt to streamline pre-operative planning. Initial phase of the study will involve model development and training through the current gold standard of virtual surgical planning. Upon conclusion of training, a new set of radiographs will be utilized to assess and compare model outcomes to surgeon annotations for determination of accuracy and efficiency.