The Role of Artificial Intelligence in Equine Colic: A Scoping Review of Diagnostic, Prognostic, and Decision-Support Applications.

Overview

• This research article provides a comprehensive review of how artificial intelligence (AI) is being used to diagnose, predict outcomes, and support clinical decisions in cases of equine colic, a serious health issue in horses.
• It summarizes various AI models, evaluates their effectiveness, and identifies methodological challenges that limit their application in veterinary practice.

Introduction to Equine Colic and AI

• Equine colic is one of the leading causes of death in horses, requiring quick and accurate clinical decisions for better outcomes.
• Artificial intelligence (AI) has emerged as a promising tool to assist veterinarians by providing diagnostic support, prognostic predictions, and aiding in treatment decisions.
• Despite growing interest, a systematic understanding of AI’s role, its methodologies, and its performance in equine colic was lacking prior to this review.

Scope and Methodology of the Review

• The review included 16 studies published between 2015 and early 2025 focused on AI applications in equine colic.
• These studies involved analysis of 104 different AI models, highlighting the diversity and volume of AI research in this field.
• The review specifically aimed to map:
  • Which AI algorithms were most commonly used
  • How these models performed based on reported metrics
  • Key methodological issues impacting practical use

Key Findings on AI Models and Their Performance

• The majority of AI models centered around prognostic predictions, especially predicting survival outcomes in horses with colic.
• Logistic regression was the most frequently employed algorithm, known for its interpretability and use in clinical settings.
• Random forest (RF) models and ensemble methods were also widely used, showing robust predictive power.
• Random forest models often achieved:
  • Area under the curve (AUC) values between 0.79 and 0.99, indicating excellent discriminative ability.
  • Prediction accuracy frequently surpassed 80%, which is promising for practical prognosis.

Methodological Challenges Identified

• Few studies performed external validation, which limits confidence in AI models’ performance across different clinical populations.
• Calibration—ensuring that predicted probabilities match observed outcomes—was rarely assessed, impacting clinical reliability.
• Standardized data preprocessing protocols were inconsistently applied, making replication and comparison between studies difficult.
• Handling of class imbalance (unequal distribution of outcome categories such as survival vs. death) was limited, potentially biasing model results.
• Reporting practices varied, reducing transparency and the reproducibility of AI findings in equine colic research.

Implications for Clinical Practice and Future Research

• The review confirms that AI has significant promise to enhance prognostic prediction and clinical decision-making in equine colic cases.
• Critical methodological barriers must be overcome before AI tools can be routinely implemented in veterinary clinics.
• Recommendations for future work include:
  • Promoting multicenter collaborations to increase data diversity and facilitate robust validation.
  • Adding external validation and rigorous assessment of calibration to verify model generalizability and reliability.
  • Establishing standardized frameworks for data preprocessing and reporting to improve study comparability.
  • Employing strategies to manage class imbalance, such as resampling techniques or customized loss functions.
  • Incorporating explainable AI methods to help veterinarians understand AI predictions, increasing trust and adoption.
  • Developing multimodal monitoring systems that integrate different types of clinical data for comprehensive decision support.

Conclusion

• This scoping review provides a detailed snapshot of the current landscape of AI applications in equine colic.
• It highlights successful approaches, cautions about important limitations, and sets a clear roadmap for advancing AI toward practical clinical use.
• Improved methodologies and collaborative approaches will be essential to harness AI’s full potential in improving outcomes for horses suffering from colic.