Machine learning techniques in biomedical data analysis: A scoping review of biostatistical perspectives

Background: Machine learning (ML) is increasingly applied in biomedical research to analyse complex datasets from imaging, omics, clinical records, and other sources.However, the extent to which biostatistical methods are integrated into ML applications remains inconsistent.Objective: This scoping review explores the use of ML techniques in biomedical research and examines the incorporation of biostatistical methods for model validation and performance evaluation.Methods: A total of 32 peer-reviewed studies published between 2018 and 2024 were reviewed.Eligible studies involved the application of ML in biomedical domains and reported statistical validation techniques or performance measures.Results: Supervised learning was the most commonly used approach, while deep learning methods were frequently employed in high-dimensional domains such as imaging and omics.Although several studies reported using cross-validation, feature selection, and standard performance metrics, integration of biostatistical frameworks was limited in scope and depth.Conclusion: The review highlights a growing trend of ML adoption in biomedical research, alongside a critical need for stronger integration of biostatistical approaches.Improved collaboration between data science and statistics is essential to ensure model reliability, interpretability, and real-world clinical applicability.