Projects
Machine Learning for Pain Intensity Estimation
Summary: Development of pain intensity estimation tool to objectively assess patients’ pain levels by leveraging data science and machine learning algorithms (physiological signals data collection, signal processing, feature extraction, support vector machine, deep learning, and logistic regression.)
The main aim is to help clinicians by accurately measuring patients’ pain levels and making the prescription process more accurate.
Network for Pain Research Literature
Summary: Development of automatic literature review methodology using graph data science algorithms (keyword co-occurrence network) to analyze more than 300,000 articles.
The main aim is to infer knowledge maps and emerging trends in the pain research literature to direct pain researchers.
Graph for Drug Development Research
Summary: Development of a search engine to extract the relationship between drugs, patients, and adverse drug reactions using graph data science and machine learning algorithms to find any potential hidden relation by link prediction methods (logistic regression, neural networks, support vector machines, and community detection.)
The main aim is to help scientists during the drug development process.
Predictive Modeling for Hospital Admission
Summary: Development of machine learning models to predict the rate of readmission (logistic regression) and length of stay of high-risk patients (linear regression) and computer simulation to analyze the complex workflows. Synchronizing real-world data (EPIC) and working with clinicians, scientists, and engineers at Mount Auburn Hospital in Boston.
​
The main aim is to reduce the readmission rate of this hospital.
Reinforcement Learning for Resource Allocation
Summary: Development of reinforcement learning-based, such as Q-learning, algorithms to simulate hospital admission of urgent and non-urgent patients.
​
The main aim is to decide the resource capacity in a dynamic healthcare environment to improve access to healthcare.
Agent-Based Modeling for Opioid Epidemic and COVID-19 Pandemic
Summary: Development of agent-based models to predict future trajectories of the opioid epidemic in the USA and the COVID-19 pandemic in the world. Application of optimization algorithms (particle swarm, genetic algorithm, and simulated annealing) to obtain the best parameter set and to analyze the interaction between agents and how they affect the spread of the epidemic.
​​
The main aim is to predict future trajectories of epidemics or pandemics to intervene early and prevent deaths.
