Machine learning based models for predicting presentation delay risk among gastric cancer patients.

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Presentation delay in gastric cancer patients leads to delayed diagnosis and treatment, negatively impacting prognosis.
• Predicting the risk of presentation delay is crucial for improving patient outcomes.
• Machine learning offers a promising approach to develop predictive models for clinical decision support.

Purpose of the Study:

• To develop and validate machine learning models for predicting presentation delay risk in gastric cancer patients.
• To identify key factors contributing to presentation delay.
• To select the optimal model for clinical application.

Main Methods:

• Utilized a derivation cohort (875 gastric cancer patients) and an external validation cohort (200 patients).
• Developed six machine learning models: logistic regression, SVM, random forest, GBDT, XGBoost, and MLP.
• Assessed model performance using metrics like AUC, accuracy, sensitivity, specificity, PPV, NPV, F1-Score, and Brier scores.

Main Results:

• The incidence of presentation delay was 39.3%.
• All developed models demonstrated good discrimination and calibration, with AUC ranging from 0.893-0.925.
• The Random Forest (RF) based model was selected as the best performer, showing strong results in both internal and external validation.

Conclusions:

• The RF-based model exhibits favorable performance for predicting presentation delay in gastric cancer patients.
• This model can assist healthcare professionals in identifying high-risk patients.
• Early identification facilitates targeted interventions to mitigate presentation delay and improve treatment outcomes.