Undergraduate Summer Internship in Health Data Research

2024 projects

• Data Processing Pipelines for Transforming 3D Optical Motion Capture Data into Wearable Sensor Data

  Host: Northumbria University – Bing Zhai (Computer and Information Sciences)

  This research proposal focuses on an innovative summer internship project designed to develop a data processing pipeline for transforming 3D optical motion capture data into wearable sensor data, including accelerometer and gyroscope outputs, for individuals with Parkinson’s disease (PwP). The project’s first objective is to construct a pipeline that converts 3D optical positional data into SMPL-based motion data utilising existing tools such as Soma and Mosh++. The second objective involves using the PIP to derive wearable sensor data from the SMPL models, enabling human motion reconstruction in a format accessible for real-world application.

  The revised final goal is to assess the existing motion synthesis algorithms’ performance in reconstructing human motion for the PwP population by identifying and understanding the error margins when these algorithms are applied to patient data. This adjustment shifts the focus from refining algorithms to a comprehensive evaluation of their current capabilities and limitations, offering valuable insights for future enhancements. By analysing clinical gait assessment data from individuals with Parkinson’s, this project seeks to bridge the gap in research that often overlooks the specific needs of this population, thereby enhancing the understanding of disease progression and fall risk assessment through improved data analysis techniques.

• Predicting Clinical Trial Success using Machine Learning

  Host: University of Newcastle – Chris Marshall (NIHR Innovation Observatory)

  Robust clinical trials are crucial for rigorously evaluating the efficacy and safety of novel health interventions., However, the premature termination of these trials can result in significant research waste, depriving health systems of crucial resource that could be better used elsewhere to enhance patient care and outcomes. Within the NIHR Innovation Observatory, we are developing an advanced machine learning approach to predict the likelihood of a trial’s success, trained using our internal OpenScan clinical trial dataset. We are excited to offer undergraduate students the opportunity to engage with this innovative work and contribute to the following objectives:

  Objectives:

  • To explore and apply feature engineering techniques to distil pertinent features from the OpenScan dataset to drive accurate predictions.
  • To support building and refinement of state-of-the-art machine learning algorithms, including regression models, tree models, deep learning models to develop predictive models.
  • To assess the interpretability of these models to identify the critical factors that drive the success rates of clinical trials and provide insights that could support future trial design.

  Supervisors: Dr Hongbo Bo, Dr Saleh Mohammed, Ross Fairbairn and Dr Christopher Marshall

• Fairness in Resource Allocation Across UK Hospitals: A Data-Driven Approach

  Host: University of Newcastle – Carlos Gonzalez (School of Computing)

  Algorithmic fairness is a key piece of Responsible Artificial Intelligence (RAI). This proposal aims to investigate the fairness of resource allocation among multiple hospitals within the UK National Health Service (NHS), to improve the overall quality of healthcare services. By leveraging NHS data sets, the aim is to identify disparities, assess the effectiveness of current allocation strategies, and to propose alternative, data-driven allocations that improve the common good. This project consists of four objectives:

  1. Evaluate the distribution of resources (staff, funding, equipment) across hospitals in the UK.
  2. Identify socio-demographic factors influencing resource allocation patterns.
  3. Assess the impact of resource allocation on healthcare outcomes and patient satisfaction.
  4. Propose data-driven recommendations to enhance the fairness and efficiency of the resource allocation processes.

  Methodology:

  Utilising anonymised NHS data sets , we will perform statistical analyses to analyse resource allocation trends. Machine learning models will be used to assess the association between demographics, resource allocation, hospital characteristics, and patient outcomes. Additionally, geographical mapping will visualise regional disparities in resource distribution. Finally, algorithmically-generated resource distributions satisfying standard fairness definitions, e.g., equality of opportunity and equality of outcomes, will be proposed.

  Significance:

  Findings will inform policymakers, healthcare administrators, and stakeholders on strategies to optimise resource allocation processes, ultimately enhancing healthcare accessibility and quality across the UK.

  Conclusion:

  This project seeks to uncover insights that result in a fairer and more effective allocation of resources among UK hospitals, thus advancing the overarching goal of equitable healthcare provision.