Privacy Protection Scheme for Detecting Double Eyelid Surgery Based on Federal Learning and Blockchain Technology

作者：Nanru Peng, Ya Han, Lifeng Ren · 发表于：International Symposium on Artificial Intelligence in Medical Sciences · 年份：2024 · DOI：10.1145/3706890.3706967 · 研究领域：Computer Science

The natural thinning or absence of upper eyelid folds results in a lack of dynamic and bright beauty in the eyes of this group of people. Reconstructing the folds on the eyelids to visually enlarge the eyes is a desire of many Asians. Double eyelid surgery has become one of the most popular shaping surgeries, with a high demand among both men and women. The success of eye shaping surgery not only depends on the superb surgical skills of shaping surgeons, but also strongly related to reasonable preoperative design. However, there is currently a lack of unified evaluation tools to assist shaping surgeons in preoperative double eyelid design. Although some objective guidance has been provided through measuring partial data in multiple studies, due to the limited sample size included in the experiments and the small number of doctors involved in the surgery, mostly from individual clinics or hospitals, the guidance provided has certain limitations. Therefore, if we want to conduct large-scale nationwide collaborative research across different practice environments, we need a large centralized database. However, centralized machine learning poses a risk of user privacy information leakage. In this study, we leverage federated learning technology to develop a collaborative model for double eyelid surgery evaluation, incorporating multiple institutions and departments. This approach ensures the protection of user privacy by preventing the risk of data leakage. Meanwhile, this paper ...