QUASR Contributes to Research on Trustworthy AI in Healthcare
Exploring the Importance of Trust, Explainability and Data Quality in Medical AI
Artificial intelligence is increasingly influencing healthcare, from medical imaging and clinical decision support to surgical planning and robotics. However, as AI becomes more integrated into medical practice, important questions remain: Can healthcare professionals trust AI-generated predictions? Are algorithms affected by bias? How can the quality and reliability of medical data be assessed?
These questions are central to the growing field of trustworthy and explainable artificial intelligence (AI). Researchers associated with the Queensland Unit for Advanced Shoulder Research (QUASR), Queensland University of Technology (QUT) and the ARC Training Centre for Joint Biomechanics contributed to a major systematic review examining these challenges.
Published in Information Fusion in August 2023, the paper, titled “A Systematic Review of Trustworthy and Explainable Artificial Intelligence in Healthcare: Assessment of Quality, Bias Risk, and Data Fusion,” examines the current research landscape and identifies important considerations for developing more reliable AI applications in healthcare.
Why Trustworthy AI Matters in Healthcare
AI systems can help clinicians analyse complex information, identify patterns and support clinical decision-making. However, their usefulness depends on the quality of the data and methods used to develop and evaluate them.
Poor-quality data, algorithmic bias and limited transparency can affect the reliability of AI-generated results. Differences in patient populations, geographical settings and clinical characteristics may also influence how well an AI model performs.
The research highlights the importance of evaluating AI systems beyond their predictive capabilities. Trustworthy healthcare AI must also consider data quality, fairness, explainability, robustness and the responsible use of information.
A Systematic Review of Healthcare AI Research
The research team examined 64 contributions from multiple academic databases, including ScienceDirect, Scopus, Web of Science and IEEE Xplore.
The review organised healthcare AI research into seven broad categories:
Explainable robotics
Prediction
Decision support
Blockchain
Transparency
Digital health
Review studies
The researchers also examined data fusion approaches, including data, feature, image, decision, multimodal, hybrid and temporal fusion. These approaches can help integrate different sources and types of healthcare information to support more comprehensive analysis.
Addressing Quality, Bias and Explainability
A central focus of the review was the challenge of establishing trust in medical AI.
The researchers explored how methodological quality and bias risks can affect the reliability of AI applications. They also examined explainable AI methods that aim to make the reasoning behind algorithmic outputs more understandable.
The paper discusses 17 explainable AI methods and identifies eight important research gaps, offering recommendations and directions for future work.
These considerations are particularly relevant in healthcare, where AI-supported outputs may influence clinical assessments and treatment decisions. Explainability and transparency are important for helping healthcare professionals understand the limitations of AI systems and assess their outputs alongside clinical evidence.
A Collaborative Research Contribution
The publication brought together researchers from several institutions and disciplines. Contributors associated with QUT and the ARC Training Centre for Joint Biomechanics included Laith Alzubaidi, Jinshuai Bai, Asma Salhi, Prof Ashish Gupta and Yuantong Gu.
The work reflects collaboration across healthcare, engineering and data science, demonstrating the importance of interdisciplinary research in addressing complex challenges in medical AI.
Prof Ashish Gupta acknowledged the contributions of the wider research team and highlighted the importance of this publication as an early contribution to the group's ongoing work in trustworthy AI.
Building a More Evidence-Based Future for Healthcare AI
AI has considerable potential to support healthcare, but its adoption must be guided by scientific evidence, reliable data and appropriate evaluation.
The review contributes to this discussion by examining the challenges of quality, bias, explainability and information fusion. Its findings provide a framework for understanding important research gaps and areas requiring further investigation.
For QUASR and its research collaborators, this work represents an important contribution to exploring how data science and AI can support the future of healthcare.
Read the Research Paper: A Systematic Review of Trustworthy and Explainable Artificial Intelligence in Healthcare
Related LinkedIn Post: View Yuantong Gu's LinkedIn post