e-ISSN: 2582-5208 International Research Journal of Modernization in Engineering Technology and Science ( Peer-Reviewed, Open Access, Fully Refereed International Journal ) Volume:07/Issue:02/February-2025 Impact Factor- 8.187 www.irjmets.com www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science [969] GENERATIVE AI MODEL FOR CHEMOTHERAPY-INDUCED MYELOSUPPRESSION IN CHILDREN Santosh Kumar *1 *1 HCL Tech, CO, USA. DOI : https://www.doi.org/10.56726/IRJMETS67323 ABSTRACT Chemotherapy-induced myelosuppression (CIM) is a major concern in pediatric oncology, often leading to severe complications, including infections, anemia, and thrombocytopenia. Predicting CIM accurately can help in optimizing treatment plans and minimizing adverse effects. This research explores the development of a Generative AI model that leverages deep learning to predict and categorize CIM in children undergoing chemotherapy. The model integrates clinical data, blood count trends, and patient history using a Transformerbased generative framework. Our ϐindings suggest that the AI model signiϐicantly improves prediction accuracy and provides actionable insights for personalized treatment adjustments. Keywords: Generative AI, Myelosuppression in Children, AI for Pediatric Health, Pediatric Oncology, Chemotherapy-Induced Myelosuppression, Myelosuppression Prediction Model. I. INTRODUCTION Chemotherapy is an essential treatment modality for various pediatric cancers, but it frequently leads to myelosuppression, a potentially life-threatening condition characterized by a decrease in bone marrow activity. Early detection and prediction of myelosuppression can enable clinicians to take proactive measures, reducing hospitalizations and improving patient outcomes. Traditional statistical models and heuristic approaches have limitations in handling complex, nonlinear patient data. Generative AI offers a novel approach by learning intricate patterns in medical data and simulating realistic scenarios for better predictive analytics. This paper details the development and validation of a Generative AI model for CIM prediction in children. II. METHODOLOGY 1. Data Collection: Clinical datasets were obtained from pediatric oncology units, including patient demographics, chemotherapy regimens, complete blood count (CBC) reports, and prior myelosuppression episodes. 2. Preprocessing: Data underwent normalization, missing value imputation, and feature engineering. 3. Model Architecture: A Transformer-based generative model was designed, leveraging a Variational Autoencoder (VAE) coupled with a Recurrent Neural Network (RNN) to capture temporal dependencies in blood cell count trends. 4. Training and Validation: The model was trained on historical patient data with stratiϐied k-fold validation to ensure robustness. 5. Evaluation Metrics: Performance was assessed using accuracy, F1-score, sensitivity, speciϐicity, and area under the receiver operating characteristic (ROC-AUC) curve. III. MODELING AND ANALYSIS 1. Model Algorithm 1.1 Data Preprocessing Dataset The dataset consists of pediatric chemotherapy patient records. It includes: • Demographics: Age, weight, gender • Chemotherapy Regimen: Type, dosage, frequency • Hematological Data (Pre & Post Treatment):