Year: 2025
Type: conference-paper
Abstract: Kabuki Syndrome (KS) is a rare genetic disorder characterized by distinctive facial features. There is an international consensus on the diagnosis of this disease, which is based on the presence or absence of multiple specific facial features of the patients. The scoring results have significant implications for the diagnosis of the disease and serve as a qualitative diagnostic criterion. For different patients with specific facial features, this paper presents a new quantitative diagnostic standard for the facial features of Kabuki Syndrome based on data analysis methods. Firstly, based on the existing diagnostic standards for Kabuki Syndrome facial features, five types of quantifiable facial features were selected. The MediaPipe Face Mesh and other methods were used to extract feature data from the obtained facial images. Secondly, a Random Forest (RF) model was constructed to study the classification and prediction of the main facial feature data of patients with Kabuki Syndrome. The importance of the main features was analyzed using the Eli5 and SHAP methods, and the weight analysis of the main facial features was obtained. Finally, a new quantitative diagnostic standard for the facial features of Kabuki Syndrome based on the importance of features was proposed. The effectiveness of the method was verified through the test set.
Authors: Ruxin Ding, Yiran Zhou, Xinyan Zong
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Funder: Soochow University
OpenAlex ID: W4413008295