Deep learning-based chest X-ray age serves as a novel biomarker for cardiovascular aging

Abstract

Chest X-ray (CXR) is one of the most commonly performed medical imaging tests. Although aging, sex and disease status have been known to cause changes in CXR findings, the extent of these effects has not been fully characterized. Here, we present a deep neural network (DNN) model trained using more than 100,000 CXRs to estimate the patient's age and sex solely from CXRs. Our DNN exhibited high performance in terms of estimating age and sex, with Pearson's correlation coefficient between the actual and estimated age of above 0.9 and an area under the ROC curve of 0.98 for sex estimation. The difference between the actual and estimated age is large in CXRs with abnormal findings, suggesting that the estimated age ("CXR age") can be a biomarker for disease status. Furthermore, by applying our DNN to CXRs of consecutive 1,562 hospitalized heart failure patients, we demonstrated that an elevated CXR age is not only associated with aging-related diseases, such as hypertension and atrial fibrillation, but also a worse outcome of heart failure. Given these results, our new concept "CXR age" serves as a novel biomarker for cardiovascular aging and can help clinicians to predict, prevent, and manage cardiovascular diseases.

Using Deep Neural Networks for Predicting Age and Sex in Healthy Adult Chest Radiographs

“…Some studies have reported age estimation from CXR images [ 39 ]. Hirotaka reported that the top of the mediastinum helps in predicting patients’ age [ 14 ]. Their results were similar to our study.…”

“…Similarly, CXR-age derived from chest radiographs may be of potential use as an imaging biomarker to represent the condition of the thorax. In fact, there have been emerging studies using CXR-age to successfully predict longevity, mortality, and cardiovascular risk [ 13 , 14 , 27 ], which provides a sound base for the imaging biomarker hypothesis.…”

“…The advantages for medical imaging to predict age and sex may have roots in the fact that age and sex are not simply one single concept. In fact, image findings may reflect the overall effects of chronological, biological, and/or pathological changes [ 13 , 14 , 15 ]. Chronological changes reflect natural development over time, e.g., endochondral ossification in skeletal evolution, changes in body shape, or breast growth related to hormones.…”

“…Among all medical imaging modalities, the chest radiograph is the most widely available modality, displaying plenty of information about the cardiopulmonary and musculoskeletal system. Despite research using chest radiograph-derived biological age estimates to successfully predict longevity, long-term mortality, and cardiovascular risk [ 13 , 14 , 27 ], very few studies have ever validated chest radiography-based age prediction methods [ 15 , 27 ], and none of them have validated the prediction model in a real-world clinical dataset. Therefore, there remains a lack of direct proof or predictive performance data regarding whether and how chest radiographs can provide accurate age and sex predictions based on clinical datasets.…”

Assessing the Performance of Automated Prediction and Ranking of Patient Age from Chest X-rays Against Clinicians

“…Patient age and gender are predicted from a proprietary dataset with a CNN in [22], with Grad-CAM [20] used to visualise characteristic features. Recent work has shown that overestimated patient age is predictive of cardiovascular and all-cause mortality rates [8,16]. GANs [5] have previously been used in a variety of medical image processing applications [23].…”