Generative Adversarial Networks-Driven Synthetic Patient Data Creation for Risk Identification in Childhood Acute Lymphoblastic Leukemia
9th International Symposium on Multidisciplinary Studies and Innovative Technologies, ISMSIT 2025, Ankara, Turkey, 14 - 16 November 2025, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/ismsit67332.2025.11268001
- City: Ankara
- Country: Turkey
- Keywords: acute lymphoblastic leukemia, artificial intelligence, childhood leukemia, generative adversarial networks, synthetic patient data
- Isparta University of Applied Sciences Affiliated: Yes
Abstract
This study aims to enhance real-world clinical data in childhood acute lymphoblastic leukemia (ALL) patients and enable more detailed patient analysis through risk category classification. New synthetic patient samples were generated via Generative Adversarial Network (GAN) model. Here, the GAN discriminator used patient data derived from a Graph Neural Network (GNN). The similarity between the synthetic patient data generated by the model and real patient data was assessed using various metrics. Distribution similarity analyses (t-SNE, PCA, and various statistical metrics) demonstrated substantial alignment between the synthetic and real patient datasets. The findings of the study demonstrated the effectiveness of GAN-based patient data augmentation in terms of data privacy and patient diversity. Furthermore, synthetically generated patient data can serve as a robust foundation for clinical digital twin applications and decision support systems.