Andrew Soltan
MB BChir MA (Cantab) MA (Oxon) MRCP
NIHR Academic Clinical Fellow & Dissertation Lead, MSc Applied Digital Health
- Specialty Registrar in Medical Oncology, Oxford University Hospitals NHS Foundation Trust
- Clinical AI Researcher
- Dissertation Lead, MSc Applied Digital Health
- NIHR Academic Clinical Fellow
I am a Specialty Registrar in Medical Oncology at Oxford University Hospitals NHS Foundation Trust, and Clinical AI Researcher and Dissertation Lead for the MSc in Applied Digital Health at the University of Oxford.
My research applies federated techniques and transformer-architecture models to routinely collected healthcare data, aiming to develop, evaluate and deploy clinical tools for Cancer care pathways.
During my NIHR Academic Clinical Fellowship, I specialised in developing AI-enabled screening and prognostic tools using routinely collected tabular data. Working with Professor David Clifton, I led development and evaluation of an AI screening test for COVID-19 in emergency departments based upon blood tests and vital signs that are collected routinely within 1h of a patient arriving in hospital. The CURIAL AI test was piloted in Oxford's John Radcliffe Hospital Emergency Department in 2021, and described in an accompanying The Lancet Digital Health commentary as "an elegant breakthrough to enhance the clinical decision-making process in the age of artificial intelligence". Complimentary work has explored model-level approaches to reduce bias in AI predictions.
To support confidential development of AI models within the NHS, I developed a new and scalable platform for end-to-end federated learning, using inexpensive micro-computing devices. The CURIAL-Federated platform was piloted across 4 NHS Trusts in 2022 to train and evaluate a Covid-19 screening test, with participating hospitals retaining custody of their data at all times, and published in The Lancet Digital Health in February 2024.
Alongside my clinical and research interests, I lead the 10-week practical dissertation module for the MSc in Applied Digital Health. The dissertation placements offer students an immersive practical experience and chance to develop a specialism within the digital health research and translational pipeline, while expanding their professional networks. Previous project placements have included opportunities to work with local hospitals, health informatics and engineering research groups, and startups within the Oxford ecosystem.
Recent publications
-
Addressing label noise for electronic health records: insights from computer vision for tabular data
Journal article
Yang J. et al, (2024), BMC Medical Informatics and Decision Making, 24
-
Mitigating machine learning bias between high income and low–middle income countries for enhanced model fairness and generalizability
Journal article
Yang J. et al, (2024), Scientific Reports, 14
-
Knowledge abstraction and filtering based federated learning over heterogeneous data views in healthcare.
Journal article
Thakur A. et al, (2024), NPJ Digit Med, 7
-
Knowledge abstraction and filtering based federated learning over heterogeneous data views in healthcare
Journal article
Thakur A. et al, (2024), npj Digital Medicine, 7
-
Generalizability assessment of AI models across hospitals in a low-middle and high income country.
Journal article
Yang J. et al, (2024), Nat Commun, 15
-
Generalizability assessment of AI models across hospitals in a low-middle and high income country
Journal article
Yang J. et al, (2024), Nature Communications, 15
-
From data to diagnosis: skin cancer image datasets for artificial intelligence
Journal article
Wen D. et al, (2024), Clinical and Experimental Dermatology, 49, 675 - 685
-
Deep reinforcement learning for multi-class imbalanced training: applications in healthcare
Journal article
Yang J. et al, (2024), Machine Learning, 113, 2655 - 2674
-
A scalable federated learning solution for secondary care using low-cost microcomputing: privacy-preserving development and evaluation of a COVID-19 screening test in UK hospitals
Journal article
Soltan AAS. et al, (2024), The Lancet Digital Health, 6, e93 - e104
-
Comparative evaluation of large-language models and purpose-built software for medical record de-identification
Preprint
Kuo R. et al, (2024)