Optimization of chemotherapy using metaheuristic optimization algorithms / Prakas Gopal Samy

Prakas Gopal , Samy (2024) Optimization of chemotherapy using metaheuristic optimization algorithms / Prakas Gopal Samy. PhD thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (227Kb)
	PDF (Thesis PhD) Restricted to Repository staff only until 31 December 2025. Download (1055Kb)

Abstract

This research undertakes a comprehensive exploration aimed at optimizing cancer therapy by integrating mathematical modelling and advanced optimization methodologies. The central focus revolves around the dual objective of minimizing tumour cell populations while optimizing chemotherapy administration to ensure healthy effector-immune cell levels. Anchored in mathematical modelling using ordinary differential equations (ODE), the study culminates in a Multi-Objective Optimal Control Problem (MOCP), categorizing into Pure MOCP (P-MOCP) and Hybrid MOCP (H-MOCP). The realization of this objective is steered by two distinct methodologies: the Method for Pure Multi-Objective Optimal Control (PMM) and the Hybrid Method (HM). Leveraging state-of-the-art Multi-Objective Optimization algorithms, the research delves into the intricate dynamics of tumour and effector cell interactions within the context of chemotherapy. The study harnesses MATLAB's ode45 solver, tailored for solving ODE. Utilizing robust Runge-Kutta methods, the solver adeptly navigates the complexities of the ODE model, particularly useful when analytical solutions are challenging to obtain. Advancing multi-objective optimization techniques for cancer treatment strategies, the study strategically incorporates Swarm Intelligence (SI) and Evolutionary Algorithms (EA). Diverging in constraints handling and Pontryagin Maximum Principle (PMP) application, the PMM and HM methodologies are evaluated through the lens of SI and EA, and their outcomes are visualized via the Pareto Optimal Front (PF). Over the simulation period, consistent trends of tumour population reduction and optimized drug administration reaffirm the methodologies' inherent effectiveness. The evaluation process employs the Hypervolume (HV) indicator and Inverted Generational Distance (IGD) to scrutinize solution efficacy and coverage. The HM emerges as a dominant strategy, driven by the Multi-Objective Differential Evolution (MODE) algorithm under literature-based control parameter settings for the mathematical model. This methodology efficiently positions solutions on the PF with minimal distances from the origin, signifying optimality. To comprehend the intricate dynamics of cancer and effector cells under chemotherapy's influence, the study undertakes equilibrium point stability analysis through the Jacobian matrix and eigenvalues. Stability and bifurcation analyses illuminate the model's responsiveness to varying control parameters. Empirical validation and performance assessment are conducted through rigorous benchmarking simulations on the PlatEMO platform, covering ten algorithms. The Multi-Objective Particle Swarm Optimizer (MOPSO) demonstrates superior performance in the HM, while the Modified Multi-Objective Particle Swarm Optimizer (M-MOPSO) excels within the PMM, highlighting its crucial role in optimizing cancer therapy with enhanced control parameters. The conclusion drawn from these analyses, spanning key metrics such as runtime, IGD, and HV, and utilizing the methodologies employed in this study, establishes the superior performance of HM in the first phase, whereas in the second phase, PMM emerges as the more effective approach. This finding underscores the HM's and PMM’s potential to revolutionize chemotherapy strategies, offering a promising avenue for advancing cancer therapy while mitigating adverse effects on healthy cell populations. Beyond demonstrating transformative potential, the research accentuates innovative methodologies' ability to shape the future of cancer treatment, instilling optimism in patients and healthcare professionals alike.

Actions (For repository staff only : Login required)