ECEESPE2025 Poster Presentations Multisystem Endocrine Disorders (43 abstracts)
Assessment of facial morphologic features in patients with polycystic ovary syndrome using deep learning:a multi-center cross-sectional study
Jiayi Lu 1 , Yucan Zhang 2,3 , Yao Wang 4 , Weihong Song 5 , Na Li 6 , Xiaowen Shao 7 , Shen Qu 8 , Yuguang Wang 3,9,10 & Manna Zhang 8,11
1Department of Clinical Laboratory, Shanghai Tenth People’s Hospital, School of Medicine, Tongji University, Shanghai, China; 2School of Mathematical Sciences and Institute of Natural Sciences, Shanghai Jiao Tong University, Shanghai, China; 3School of Mathematics and Statistics, University of New South Wales, New South Wales, Australia; 4Department of Dermatology, Shanghai Tenth People’s Hospital, Tongji University School of Medicine, Shanghai, China; 5Department of Endocrinology, The first People’s Hospital of Chenzhou, Chenzhou, China; 6Department of Endocrinology and Metabolism, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; 7Department of Obstetrics and Gynecology, Shanghai Tenth People’s Hospital, Tongji University School of Medicine, Shanghai, China; 8Department of Endocrinology and Metabolism, Shanghai Tenth People’s Hospital, School of Medicine, Tongji University, Shanghai, China; 9Institute of Natural Sciences and School of Mathematical Sciences, Shanghai Jiao Tong University, Shanghai, China; 10Shanghai AI Laboratory, Shanghai, China; 11Department of General Practice, Shanghai Tenth People’s Hospital, School of Medicine, Tongji University, Shanghai, China
JOINT2146
Aim: To investigate the application of deep learning-based facial recognition methods to facilitate the diagnosis of PCOS by leveraging the distinctive phenotypic features associated with the condition.
Methods: A muti-center, cross-sectional study was conducted from three tertiary hospitals in China from June 2023 to August 2024. A total of 163 participants with polycystic ovary syndrome (PCOS) and 162 non-PCOS women were recruited. Images were captured from multiple angles and were supplemented with clinical and metabolic data, including body mass index (BMI), glycated hemoglobin (HbA1c), blood lipid profiles, and sex hormone levels. Three convolutional neural network architectures (VGG-Net, ResNet-50, Inception-ResNet-v2) and Gradient-weighted Class Activation Mapping (Grad-CAM) were employed to distinct discriminative features in patients with PCOS.
Results: The mean age in participant with PCOS was 26.56 years old and have a higher body mass index (BMI), reduced number of menstrual cycles per year, elevated total testosterone and worse glycolipid metabolism than those without PCOS. Inception-ResNet-v2 achieved the highest accuracy for PCOS diagnosis, at 82.1%, whereas ResNet-50 and VGG-16, with an accuracy of 78.57% and 73.21% respectively. Inception-ResNet-v2 (AUC = 0.886) demonstrates the best performance for PCOS diagnosis, with its curve furthest from the diagonal. Additionally, the study employed Grad-CAM to visualize the model’s focus on specific facial regions, particularly the jawline, nose, and forehead, which are indicative of PCOS traits such as hirsutism and acne.
Conclusions: Our findings demonstrated facial morphologic features in patients with PCOS was distinct and Artificial intelligence (AI)-based PCOS detection could achieve satisfactory sensitivity for detecting the patients with PCOS, which suggested the feasibility and potential of AI-driven facial recognition as a non-invasive and efficient tool for PCOS screening.
Keywords: polycystic ovary syndrome, facial recognition, deep learning.
Volume 110
Article tools
My recent searches
My recently viewed abstracts
Authors
Endocrine Abstracts
ISSN 1470-3947 (print) | ISSN 1479-6848 (online)
© Bioscientifica 2025 |
Privacy policy |
Cookie settings
BiosciAbstracts
Bioscientifica Abstracts is the gateway to a series of products that provide a permanent, citable record of abstracts for biomedical and life science conferences.
Find out more