Arsene Fitzgerald Ferguson

Mr. Arsene Fitzgerald Ferguson


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 June 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
Georgios P. GeorghiouGeorgios P. Georghiou
Georgios P. Georghiou
AFArsene Fitzgerald Ferguson
Arsene Fitzgerald Ferguson
SKSotiris Kyriakou
Sotiris Kyriakou
AGAmalia Georgiou
Amalia Georgiou
AKArgyris Kyriakou
Argyris Kyriakou
FTFilippos Triposkiadis
Filippos Triposkiadis
Artificial intelligence (AI) and machine learning (ML) are increasingly being evaluated as decision-support tools for forecasting postoperative complications after cardiothoracic surgery. This narrative review synthesises peer-reviewed clinical, biomedical informatics, and digital-health literature across acute kidney injury (AKI), postoperative atrial fibrillation (POAF), pulmonary complications, neurological outcomes, mortality, wound monitoring, wearable surveillance, and post-discharge monitoring. Rather than treating reported AUC values as interchangeable evidence of algorithmic superiority, this review emphasises that model performance depends on at least three interacting factors: algorithmic architecture, data richness, and validation strategy. Conventional tools such as EuroSCORE II and Society of Thoracic Surgeons (STS) models remain clinically useful because they are familiar, transparent, and externally established; however, many rely on static perioperative snapshots and are less suited to continuously updated intraoperative and postoperative data streams. Contemporary ML models, including gradient-boosting approaches, support-vector methods, convolutional neural networks, and wearable-based algorithms, may improve discrimination in selected settings, particularly when dynamic laboratory trajectories, physiological time-series, imaging data, or remote-monitoring signals are available. However, current evidence remains limited by single-centre development, heterogeneous endpoints, inconsistent calibration reporting, limited decision-curve analysis, low positive predictive values in surveillance applications, and sparse prospective implementation. Future work should prioritise multicentre external validation, calibration, clinical utility analysis, standards-based electronic medical record interoperability, and implementation pathways that preserve clinician oversight.