HRX 2024 AbstracX Library-AI Models: Insights into Age Prediction

AI Models: Insights into Age Prediction

Pdf Summary

The research, led by Dr. Hadi Younes, explores the capability of artificial intelligence (AI) models to predict a person’s age using a two-lead electrocardiogram (ECG). This study evaluates whether simplified ECGs can provide sufficient data for accurate age prediction, which could have potential applications in identifying various cardiovascular issues.<br /><br />The AI model was developed using a large pre-training dataset called CODE, consisting of 345,779 ECGs with patients aged between 18 and 80, averaging 53.24 years old. This was later fine-tuned using the Physionet database, which contained 36,605 ECGs, of which 5,422 were normal ECGs. The study emphasized creating models for different ECG lead configurations: 12-lead, 3-lead, and notably, the 2-lead configuration. <br /><br />The results of the study indicated that the AI’s predicted age often differed from the chronological age. For example, a model with two leads could underestimate a person's age by an average of 4.5 years. The research also identified correlations between certain heart conditions and discrepancies in age prediction. For instance, when the AI-predicted age exceeded the chronological age by seven years or more, there was a higher incidence of conditions such as hypertension, low ejection fraction, myocardial infarction, coronary artery disease (CAD), and atrial fibrillation (AFib).<br /><br />In conclusion, despite using only two leads, the AI model effectively estimated the age gap between predicted and chronological age. This study suggests that even limited ECG data could be significant for early detection and management of cardiovascular conditions, offering a promising tool for enhancing diagnostic processes in healthcare.

Keywords

artificial intelligence

age prediction

electrocardiogram

cardiovascular issues

CODE dataset

Physionet database

heart conditions

diagnostic tool

healthcare

age estimation