[Paper Review] Modelling of Sickle Cell Anemia Patients Response to Hydroxyurea using Artificial Neural Networks

Brendan E. Odigwe, Jesuloluwa S. Eyitayo+2 more|arXiv (Cornell University)|Nov 25, 2019

Machine Learning in Healthcare20 references4 citations

TL;DR

This study develops a deep artificial neural network (DNN) model to predict hydroxyurea (HU) response in sickle cell anemia patients using 22 blood-derived parameters from 122 patients. The model achieves 92.6% accuracy in forecasting hemoglobin F (HbF) levels after HU therapy, enabling personalized treatment decisions and reducing unnecessary exposure for non-responders.

ABSTRACT

Hydroxyurea (HU) has been shown to be effective in alleviating the symptoms of Sickle Cell Anemia disease. While Hydroxyurea reduces the complications associated with Sickle Cell Anemia in some patients, others do not benefit from this drug and experience deleterious effects since it is also a chemotherapeutic agent. Therefore, to whom, should the administration of HU be considered as a viable option, is the main question asked by the responsible physician. We address this question by developing modeling techniques that can predict a patient's response to HU and therefore spare the non-responsive patients from the unnecessary effects of HU on the values of 22 parameters that can be obtained from blood samples in 122 patients. Using this data, we developed Deep Artificial Neural Network models that can predict with 92.6% accuracy, the final HbF value of a subject after undergoing HU therapy. Our current studies are focussing on forecasting a patient's HbF response, 30 days ahead of time.

Motivation & Objective

• To identify predictive biomarkers for hydroxyurea (HU) response in sickle cell anemia patients.
• To reduce adverse effects in non-responders by predicting HU efficacy before treatment.
• To develop a machine learning model capable of forecasting HbF levels post-HU therapy with high accuracy.
• To enable personalized medicine in sickle cell anemia by forecasting HbF response 30 days in advance.
• To utilize 22 routinely measured blood parameters as input features for clinical feasibility.

Proposed method

• A deep artificial neural network (DNN) was trained on clinical data from 122 sickle cell anemia patients.
• Input features consisted of 22 blood parameters collected prior to hydroxyurea therapy.
• The model was optimized to predict the final hemoglobin F (HbF) level after HU treatment.
• Model performance was evaluated using 10-fold cross-validation to ensure robustness.
• The architecture was designed to handle non-linear relationships between baseline parameters and HbF response.
• The study focuses on extending the model to forecast HbF levels 30 days in advance.

Experimental results

Research questions

• RQ1 Which baseline blood parameters most strongly predict HbF response to hydroxyurea in sickle cell anemia patients?
• RQ2 Can a deep neural network accurately forecast HbF levels after hydroxyurea therapy using only routine blood tests?
• RQ3 How can machine learning models reduce the risk of administering hydroxyurea to non-responding patients?
• RQ4 Can the model predict HbF response 30 days prior to treatment initiation?
• RQ5 What is the optimal model architecture for predicting HbF response using clinical biomarkers?

Key findings

• The deep neural network model achieved a prediction accuracy of 92.6% for final hemoglobin F (HbF) levels after hydroxyurea therapy.
• The model successfully identified key baseline blood parameters predictive of HU response from a set of 22 routinely measured variables.
• The model demonstrated high generalization performance through 10-fold cross-validation.
• The study confirms that machine learning can effectively forecast HbF response using only pre-treatment blood data.
• The current framework is being extended to predict HbF response 30 days in advance for clinical decision support.
• The approach offers a viable pathway toward personalized treatment planning in sickle cell anemia.