AI in healthcare: Predicting surgical outcomes for chronic exertional compartment syndrome
Researchers at Loughborough University, led by Dr Georgina Cosma, are undertaking research to identify a set of key clinical features contributing to the occupational outcomes of surgery.
Chronic exertional compartment syndrome (CECS) is a condition occurring most frequently in the lower limbs and often requires corrective surgery to alleviate symptoms. Among military personnel, the success rates of this surgery can be as low as 20%, presenting a challenge in determining whether surgery is worthwhile.
AI has the potential to transform many aspects of healthcare delivery and it can support improvements in care outcomes, patient experience and access to healthcare services. Importantly, AI technologies must be adopted in an ethical and patient-centred way, and we must proactively address risks and biases in AI-based systems to create healthcare systems that work fairly for everyone.
Research in focus
Dr Georgina Cosma and her colleagues analysed the data of 132 fasciotomies for CECS, coupled with input from clinicians, to identify a set of key clinical features contributing to the occupational outcomes of surgery.
Features were utilised to develop a machine learning model for predicting return-to-work outcomes 12-months post-surgery. An AUC of 0.85 ± 0.08 was achieved using a linear-SVM, trained using 6 features (height, mean arterial pressure, pre-surgical score on the exercise-induced leg pain questionnaire, time from initial presentation to surgery, and whether a patient had received a prior surgery for CECS).
To facilitate trust and transparency, interrogation strategies were used to identify reasons why certain patients were misclassified, using instance hardness measures. Model interrogation revealed that patient difficulty was associated with an overlap in the clinical characteristics of surgical outcomes, which was best handled by XGBoost and SVM-based models.
The methodology was compiled into a machine learning framework, termed AITIA, which can be applied to other clinical problems. AITIA extends the typical machine learning pipeline, integrating the proposed interrogation strategy, allowing the user to reason and decide whether to trust the developed model based on the sensibility of its decision-making.
- Houston A, Cosma G, Turner P, Bennett A. (2021). Predicting surgical outcomes for chronic exertional compartment syndrome using a machine learning framework with embedded trust by interrogation strategies. Scientific Reports, 11(1), 24281. DOI: 10.1038/s41598-021-03825-4
Related research
- Cosma G, McArdle SE, Foulds GA, Hood SP, Reeder S, Johnson C, Khan MA, Pockley AG. (2021). Prostate cancer: early detection and assessing clinical risk using deep machine learning of high dimensional peripheral blood flow cytometric phenotyping data. Frontiers in Immunology, 12, 786828. DOI: 10.3389/fimmu.2021.786828
- Salesi S, Cosma G. (2021). Generalisation Power Analysis for finding a stable set of features using evolutionary computation feature selection algorithms. Knowledge-Based Systems, 231, 107450, ISSN: 0950-7051. DOI: 10.1016/j.knosys.2021.107450
- Salesi S, Cosma G, Mavrovouniotis M. (2021). TAGA: Tabu Asexual Genetic Algorithm embedded in a filter/filter feature selection approach for high-dimensional data. Information Sciences, 565, pp.105-127, ISSN: 0020-0255. DOI: 10.1016/j.ins.2021.01.020
- Hood SP, Cosma G, Foulds GA, Johnson C, Reeder S, McArdle SE, Khan MA, Pockley AG. (2020). Identifying prostate cancer and its clinical risk in asymptomatic men using machine learning of high dimensional peripheral blood flow cytometric natural killer cell subset phenotyping data. eLife, 9, e50936, ISSN: 2050-084X. DOI: 10.7554/eLife.50936
- Cosma G, Mcginnity TM. (2019). Feature extraction and classification using leading eigenvectors: Applications to biomedical and multi-modal mHealth data. IEEE Access, 7, pp.107400-107412. DOI: 10.1109/access.2019.2932868
- Cosma G, McArdle SE, Reeder S, Foulds GA, Hood S, Khan M, Pockley AG. (2017). Identifying the presence of prostate cancer in individuals with PSA levels <20 ng ml−1 using computational data extraction analysis of high dimensional peripheral blood flow cytometric phenotyping data. Frontiers in Immunology, 8(DEC), ARTN 1771, ISSN: 1664-3224. DOI: 10.3389/fimmu.2017.01771
Meet the expert
Dr Cosma is a Senior Lecturer (Associate Professor) in Data Science and Artificial Intelligence at the Department of Computer Science, School of Science. Her expertise resides in natural language processing, intelligent and neural information retrieval, multi-modal machine and deep learning applied to a variety of AI applications with a particular focus on healthcare. She is particularly interested in the analysis of healthcare data (including multi-modal data) and the development of AI models that are ethical, unbiased, and which can provide reasoning behind predictions.
Dr Cosma is currently PI or CI in healthcare research projects which are funded by the NIHR and the Health Foundation, leading the development of healthcare data analysis, modelling, clustering, and classification models. Dr Cosma's research is also focused on the development of combinatorial feature selection algorithms for biomarker discovery.
She is happy to hear from those wishing to collaborate on research focussing on the analysis of healthcare records, or those looking to complete a PhD is a related area. Find out more about the current PhD students supervised by Dr Cosma.
