Chemotherapy optimization and patient model parameter estimation based on noisy measurements

Abstract

The application of the achievements of mathematics and informatics greatly helped the devel- opment of medicine. Designing personalized therapies using different algorithms is crucial, especially during chemotherapy, to minimize the toxic effects on the patient and avoid resis- tance, thus ensuring a higher quality of life. In this work, we present an LSTM neural network that can quickly and accurately estimate the parameters of the tumor dynamics model based on noisy virtual patient data. In addition, we present a genetic algorithm designed for ther- apy optimization, which is able to predict the most appropriate personalized therapy based on the estimated parameters. In this work, we focus on finding the optimal hyperparameters of this genetic algorithm. Optimizing the hyperparameters is of fundamental importance in designing the best possible personalized therapy.