Comparative Evaluation of Brain Biological Age Estimation Methods: Data Sources and Performance Benchmarks in Diverse Deep Learning Architectures

Introduction: Biological age (BA) offers a more accurate measure of an individual's health status and aging rate than chronological age. This study provides a systematic review and comparative analysis of deep learning (DL) methodologies for BA estimation.Materials and Methods: We analyzed ۳۳ selected studies, extracting data into structured tables to compare data sources (brain MRI, X-rays, blood biomarkers, wearable sensors), model architectures (CNNs, LSTMs, Ensembles, Multimodal), and performance metrics (MAE, R², AUROC). This framework enabled a transparent, side-by-side evaluation of each approach's strengths and limitations.Results: Our analysis confirms the superiority of advanced DL architectures. CNNs demonstrated exceptional performance on imaging data, with a lightweight SFCN model for brain MRI achieving a state-of-the-art Mean Absolute Error (MAE) of ۲.۱۴ years. Models that combined multiple data types, such as imaging with clinical information, proved to be the most robust. For instance, one multimodal ensemble model achieved an AUROC of ۰.۸۹-۰.۹۱ for predicting mortality. However, significant challenges were consistently identified, including limited model generalizability across diverse populations and the critical issue of data heterogeneity.Conclusion: Deep learning holds immense promise for accurate biological age estimation, with complex, data-specific models like CNNs and multimodal ensembles delivering the highest performance. For successful translation into clinical practice, future efforts must prioritize overcoming barriers related to model generalizability, data standardization, and interpretability. Resolving these issues is essential for BA to realize its potential in personalized preventive medicine and health risk assessment.

کلیدواژه ها:

Biological Age Estimation ، Deep Learning ، Neural Networks ، Artificial Intelligence