Skip to Main Content

An official website of the United States government

Investigating Survival Dynamics in Oncological Patients Through Machine Learning and Deep Learning Approaches

Principal Investigator
Name
Yukti Tripathi
Degrees
B.Tech + M.Tech Biotechnology
Institution
Amity University
Position Title
Student
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-1386
Initial CDAS Request Approval
Feb 6, 2025
Title
Investigating Survival Dynamics in Oncological Patients Through Machine Learning and Deep Learning Approaches
Summary
This project explores the use of Machine Learning (ML) and Deep Learning (DL) techniques to analyze survival dynamics in oncological patients, aiming to enhance prognosis prediction and personalized treatment strategies. By using clinical data, the study develops predictive models using algorithms such as Random Forest, XGBoost, Neural Networks, and deep survival models. Advanced survival analysis techniques, including Kaplan-Meier curves and Cox Proportional Hazards models, are employed to assess patient outcomes. Additionally, interpretability methods ensure that the predictions are clinically meaningful and useful for oncologists. Ultimately, this research contributes to precision oncology, improving survival predictions and aiding in more effective treatment planning.
Aims

Specific Aims of the Project
1. Develop Machine Learning and Deep Learning Models for Survival Prediction
• Utilize algorithms like Random Forest, XGBoost, Neural Networks, and Deep Survival Models to predict oncological patient survival.
• Optimize model performance through feature selection, hyperparameter tuning, and cross-validation.
2. Integrate Multi-Modal Data for Enhanced Prognostic Accuracy
• Combine clinical, genomic, and imaging data to improve predictive capabilities.
• Address challenges such as missing data, feature selection, and data harmonization to ensure model robustness.
3. Apply Advanced Survival Analysis Techniques
• Implement Kaplan-Meier estimators, Cox Proportional Hazards models, and deep survival networks for precise survival estimation.
• Utilize time-to-event modeling and risk stratification to better understand patient survival trends.
4. Ensure Model Interpretability and Clinical Applicability
• Use explainability techniques like SHAP and LIME to make model predictions understandable for oncologists.
• Develop a Clinical Decision Support System (CDSS) to assist in personalized treatment planning and patient management.

These aims collectively contribute to advancing precision oncology by leveraging AI to improve survival predictions and guide effective treatment strategies.

Collaborators

Dr. Hina Bansal