Research Projects
Research Projects
My primary area of study is Mathematical Oncology, focusing on the modeling and analysis of cancer dynamics through Reaction-Diffusion Equations. My current work uses physics-informed strategies, but my future work aims to utilize methods such as Physics Informed Neural Networks to uncover underlying mechanisms and predict tumor behavior. This work aims to optimize cancer diagnosis, treatment strategies and improve therapeutic outcomes. More information about my previous and ongoing research projects are provided below.
Current Projects
A Comparative Study of Machine Learning and Physics-Informed Neural Networks for Tumor Growth Modeling
Summary:
This project investigates the performance of traditional machine learning methods and Physics-Informed Neural Networks (PINNs) in modeling and predicting tumor growth. The study compares data-driven approaches, which learn tumor growth patterns from observed data, with physics-informed approaches, which incorporate mathematical equations and biological constraints into the learning process. By evaluating predictive accuracy, data requirements, and the ability to capture tumor growth dynamics, this research aims to identify the strengths and limitations of each approach. The findings may contribute to the development of reliable computational tools for understanding tumor progression and supporting future cancer treatment planning.
Key Highlights:
Comparative evaluation of traditional machine learning models and Physics-Informed Neural Networks for tumor growth prediction.
Integration of mathematical tumor growth models, such as logistic growth equations, into physics-informed learning frameworks.
Assessment of predictive accuracy, data efficiency, and model generalization.
Investigation of the role of biological constraints and governing equations in improving tumor growth predictions.
Exploration of the potential applications of machine learning and PINNs in personalized cancer modeling.
Tools & Techniques:
Python, MATLAB, NumPy, SciPy, scikit-learn, PyTorch or TensorFlow, Physics-Informed Neural Networks (PINNs), numerical simulations, differential equations, and statistical performance evaluation.
Research Methodology:
The study will develop and evaluate selected machine learning models, such as XGBoost, random forests, Classification and Regression Tree (CART) or neural networks, alongside Physics-Informed Neural Networks. Synthetic or publicly available tumor growth data will be used to train and test the models. Performance will be evaluated using metrics such as root mean squared error (RMSE), R-squared and parameter settings.
Expected Outcomes:
The project is expected to provide a systematic comparison of data-driven and physics-informed approaches for tumor growth modeling. It aims to determine when incorporating biological knowledge into neural networks improves predictive performance, reduces dependence on training data, or enhances model interpretability. The results will help identify the advantages and limitations of each approach and establish a foundation for more reliable computational models of tumor progression.
Completed Projects
Title: Reaction-diffusion modeling of vascular tumor growth: Bifurcation, relapse, and therapy efficacy
Summary: This study explores the effectiveness of chemotherapy and anti-angiogenic therapy in controlling cancer growth using an extended Reaction-Diffusion (RD) model. Incorporating cell growth dynamics and metastasis through angiogenesis, the model refines the original framework by Pinho et al. Numerical simulations, assess tumor dynamics and therapeutic outcomes, offering insights into stability, bifurcation behavior, and optimized treatments.
This project is a published work with collaborators Dr. Bani-Yagoub and Dr. Bi-Botti Youan https://doi.org/10.3934/mbe.2025109
Key Highlights:
Incorporation of extended logistic growth models.
Reaction-diffusion dynamics for metastasis and angiogenesis.
Stability and bifurcation analysis to identify treatment thresholds.
Tools & Techniques: MATLAB PDE solver, PDEPE, numerical simulations, bifurcation analysis, mathematical modeling.
Collaborators: Dr. Bani-Yagoub, Dr. Bi-Botti Youan.
Outcome: Presented in the 9th Annual Meeting of SIAM Central States Section https://sse.umkc.edu/siam-2024/
Title: Effectiveness of Chemotherapy and Anti-Angiogenic Therapy in Controlling Cancer Growth
Summary: This study investigates the effectiveness of chemotherapy and anti-angiogenic therapy in controlling cancer growth through an extended mathematical model. Numerical simulations, utilizing MATLAB’s ODE45 and the Runge-Kutta method, explore how healthy cell growth behaviors influence vascular tumor dynamics under varying initial conditions. The findings provide insights into therapeutic strategies and tumor growth patterns, contributing to ongoing advancements in cancer modeling.
Key Highlights: Modeled cell growth dynamics.
Tools & Techniques: Runge-Kutta method, MATLAB ODE 45 solver.
Outcome: Presented at the 10th Annual UMKC Math and Stat Research Day. https://sites.google.com/view/mathrd/home
Future Directions
Future work will extend the comparative study by incorporating more realistic biological processes and exploring the potential of physics-informed learning for clinical applications.
Integration of treatment effects: Extend the models to account for chemotherapy, immmunotherapy, and anti-angiogenic therapy to investigate their effects on tumor growth and treatment response.
Multispecies tumor modeling: Incorporate interactions among tumor cells, healthy cells, immune cells, and blood vessels to capture more complex tumor microenvironments.
Patient-specific tumor prediction: Investigate the use of medical imaging data and patient-specific measurements to develop individualized tumor growth models.
Hybrid modeling approaches: Develop hybrid frameworks that combine machine learning flexibility with mathematical tumor growth equations to improve predictive accuracy and biological consistency.
Clinical validation and interpretability: Assess model performance using real-world clinical or medical imaging datasets, when available, and investigate whether the resulting predictions can support treatment planning.
Long-Term Goal:
The long-term goal is to develop accurate, computationally efficient, and biologically informed tumor growth prediction models that bridge mathematical oncology, machine learning, and personalized medicine. By identifying the conditions under which conventional machine learning or PINNs perform best, this research may provide a foundation for future treatment optimization and patient-specific cancer modeling.
Publications & Presentations
Priscilla Owusu Sekyere, Majid Bani-Yaghoub, Bi-Botti C. Youan. Reaction-diffusion modeling of vascular tumor growth: Bifurcation, relapse, and therapy efficacy[J]. Mathematical Biosciences and Engineering, 2025, 22(11): 2944-2987. doi: 10.3934/mbe.2025109
Presented a poster on the published work at the BioNexus Conference in April 2026
Presented at the 2ND SIAM-CSS Meeting at Arkansas Fayeville April 2nd 2026
Presented at the AMS sectional meeting at Fargo, North Dakota April 2026
Ptresented at the Annual UMKC Math and Stat Research Day 2025
The 9th Annual Meeting of SIAM Central States Section https://sse.umkc.edu/siam-2024/
The 10th Annual UMKC Math and Stat Research Day. https://sites.google.com/view/mathrd/home