Deep Learning-Based Predictive Analysis for Neuroblastoma Using ANN-LSTM on Davide Chicco's Neuroblastoma EHRs Dataset - A Study on Using Multiple Datasets to Improve Neuroblastoma Prediction
Pravalika Boddupally, C. Kishor Kumar Reddy, S. M., Jagadeshwari Puttanapura, Anusha Adula, Sai Pranavi Alampally · 2025
Neuroblastoma is a common and lethal child cancer that, when diagnosed late, causes early death. Prognosis in the early phase is vital for facilitating timely treatment, individualized therapeutic protocols, and enhanced survival rates. This paper provides comparative evaluation of the following algorithms—Proposed Deep Learning Model (ANN + LSTM), Convolutional Neural Network (CNN), Random Forest (RF), Gradient Boosting Machine (GBM), Support Vector Machine (SVM), and Logistic Regression (LR)—to forecast neuroblastoma prognosis based on clinical, genetic, and biochemical information. Following feature selection and preprocessing, the ANN + LSTM model produced the highest results with 96.23% Accuracy, 96.8% Precision, 0.95 ROC-AUC, F1-score of 93.9%, and Error Rate of 3.77%. This study demonstrates the capabilities of AI based predictive models in enhancing pediatric cancer prognosis, helping oncologists with early diagnosis and decision-making. This research is in consonance with Sustainable Development Goal 3, Good Health and Well-being, by using AI-based models for improving disease prediction and treatment and thus relieving the healthcare system burden through early treatment.