Explainable, Fair, Reproducible, and Collaborative Surgical Artificial Intelligence: Integrating Data, Algorithms, and Clinical Reasoning for Surgical Risk Assessment
The average American can expect to undergo seven surgical operations during a lifetime. Each year in the United States, 150,000 surgical patients die and 1.5 million develop a complication after surgery. The risk for postoperative complications arises from the interaction between preoperative health and the complexity of care that the patient receives. Implementing preventive strategies that mitigate this risk depends on timely and accurate surgical risk surveillance throughout the continuum of perioperative care.
So far, the IDEALIST project has yielded 98 publications, 3 patents, and the successful implementation of a real-time intelligent surgical risk assessment system (MySurgeryRisk) at the University of Florida. Progress in medical artificial intelligence (AI), however, remains halted by limited datasets and models with insufficient interpretability, transparency, fairness, and reproducibility that are difficult to implement and share across institutions.
The next objective of the IDEALIST project is to develop a new conceptual framework for “Explainable, Fair, Reproducible, and Collaborative Medical AI” in order to provide a foundation for clinical implementation at scale. IDEALIST leverages OneFlorida, a large clinical consortium of 22 hospitals serving 10 million patients in Florida, the nation’s third largest state. To achieve the next project objective, we are pursuing three specific aims.
- External and prospective validation of a novel interpretable, dynamic, actionable, fair, and reproducible algorithmic toolkit for real-time surgical risk surveillance.
- Development and evaluation of an explainable AI platform (XAI-IDEALIST) for real-time surgical risk surveillance using human-grounded benchmarks.
- Implementation and evaluation of a federated-learning approach with advanced privacy features for collaborative surgical risk model training.
How could IDEALIST help patients?
IDEALIST could advance the ability of healthcare providers to diagnose diseases by computing the risk of postoperative complications in real-time. Ultimately, the results are expected to improve patient outcomes, decrease hospitalization costs, and reduce lifelong complications. The proposed research is relevant to public health because it can result in enhanced perioperative care workflow and early intervention.
Why is IDEALIST important to researchers?
The IDEALIST approach is innovative because it represents the first attempts to build
- the first surgical FAIR (Findable, Accessible, Interoperable, Reproducible) AI-ready, large multicenter multimodal dataset,
- novel computational approaches accompanied by assessing fairness and reproducibility,
- a multifaceted and full-stack explainable AI framework, and
- a federated learning capacity for privacy-preserving model training across institutions.
The proposed research is significant since it will address several key problems and critical barriers, including the lack of
- AI-ready large surgical datasets,
- interpretable, dynamic, actionable, fair and reproducible surgical risk algorithms,
- a medical AI explainability platform, and
- a systematic approach for collaborative model training and sharing across institutions.
OneFlorida+ Data Trust
The Data Trust is regularly updated with the inclusion of new partners and data refreshes from existing partners. All data are cleaned, transformed, curated and contained in a centralized data warehouse, allowing streamlined inquiries and uniform results based on high-quality data. Legal agreements for data use and use of OneFlorida+’s centralized IRB have already been negotiated with all partners, reducing paperwork and administrative burden.
Select IDEALIST Publications
- “Building an automated, machine learning–enabled platform for predicting post-operative complications,” Balch JA, Ruppert MM, Shickel B, Ozrazgat-Baslanti T, Tighe PJ, Efron PA, Upchurch Jr. GR, Rashidi P, Bihorac A, Loftus TJ. Physiological Measurement. 2023;44:024001. Published 2023 February 9.
- “Performance of a Machine Learning Algorithm Using Electronic Health Record Data to Predict Postoperative Complications and Report on a Mobile Platform,” Ren Y, Loftus TJ, Datta S, Ruppert MM, Guan Z, Miao S, Shickel B, Feng Z, Giordano C, Upchurch Jr. GR, Rashidi P, Ozrazgat-Baslanti T, Bihorac A. JAMA Network Open. 2022;5(5):e2211973. Published 2022 May 16.
News about IDEALIST
- “Why Hospitals Still Make Serious Medical Errors—and How They Are Trying to Reduce Them,” Laura Lando, Wall Street Journal, March 12, 2023.
- “UF researchers develop artificial intelligence to help predict complications from surgery,” Florida Physician (UF Health), June 14, 2022
- “Researchers at UF say their Artificial Intelligence system can help predict surgical problems,” WCJB 20 News (ABC), June 2, 2022.
More IDEALIST Information
- Explainable, Fair, Reproducible and Collaborative Surgical Artificial Intelligence: Integrating data, algorithms and clinical reasoning for surgical risk assessment (XAI-IDEALIST), National Institutes of Health (NIH) Reporter
Idealist dataset can be found here