Cancer Risk Prediction Using Machine Learning for Supporting Early Cancer Diagnosis in Symptomatic Patients: A...

Area of Science:

• Medical Informatics
• Oncology
• Machine Learning

Background:

• Predictive models can aid clinicians in identifying patients who may benefit from cancer investigations.
• Machine learning (ML) models are being developed to estimate cancer risk based on symptoms and patient characteristics.

Purpose of the Study:

• To systematically review published evidence on machine learning models for cancer risk prediction.
• To assess the diagnostic performance and quality of ML models using symptoms and patient characteristics.

Main Methods:

• Systematic review of studies published between 2014-2024 from MEDLINE, Scopus, and EMBASE.
• Assessment of study quality using QUADAS-AI tools and adherence to TRIPOD guidelines.
• Quantitative synthesis of diagnostic performance metrics, including accuracy, sensitivity, specificity, and AUC.

Main Results:

• 34 studies met inclusion criteria, focusing on lung, mesothelioma, and gastrointestinal cancers.
• ML models incorporated signs/symptoms, sociodemographic, and lifestyle factors, with variable performance (AUC 0.60-1).
• Most studies (94.1%) had a high risk of bias, and only 70% performed internal validation.

Conclusions:

• ML models show potential for cancer risk prediction by managing complex data.
• Heterogeneous evidence, bias, and incomplete reporting limit current ML models for clinical use.
• Further validation and real-world performance assessments are crucial for reliable clinical application.