Personalized Quantification of Facial Normality using Artificial Intelligence

Abstract

While congenital facial deformities are not rare, and surgeons typically perform operations to improve these deformities, currently the success of the surgical reconstruction operations can only be “measured” subjectively by surgeons and specialists. No efficient objective mechanisms of comparing the outcomes of plastic reconstruction surgeries or the progress of different surgery techniques exist presently. The aim of this research project is to develop an efficient software application that can be used by plastic surgeons as an objective measurement tool for the success of an operation. The long-term vision is to develop a software application that is user-friendly and can be downloaded on a regular laptop and used by doctors and patients to assess the progress of their surgical reconstruction procedures. The application would work by first scanning a face before and after an operation and providing the surgeon with a normality score of the face from 0 to 3 where 3 represents normal and 0 represents extreme abnormality. A score will be given when the face is scanned before and after surgery. The difference between those scores is what we will call the delta. A high delta value would point to a high improvement in the normality of a face post-surgery, and a low delta value would indicate a small improvement. The first chapter of the thesis represents the introduction which describes the general aspects of the project. The second chapter presents the methodology employed for building the application and the existing solutions and proposed functional model structure. The results chapter presents the process behind collecting and labeling the image database and analyzes the scores produced by the program when fed with new images from the database. Finally, the last chapter of this thesis presents the conclusions. The list of references completes this work.