Skip to Main Content

An official website of the United States government

Principal Investigator
Name
Yasuhito Mitsuyama
Degrees
M.D.
Institution
Osaka Metropolitan University
Position Title
Student
Email
About this CDAS Project
Study
PLCO (Learn more about this study)
Project ID
PLCOI-1285
Initial CDAS Request Approval
Aug 8, 2023
Title
Development of biomarkers of aging in chest x-ray images using deep learning
Summary
With an aging global population, the significance of the fields of aging and senescence is becoming more pronounced. Despite sharing the same chronological age, some individuals may require assistance in daily living, while others live independently and healthily into very old age. In addition to biological age, various biomarkers have been proposed as indicators of individual aging. Chest X-rays are a simple and common examination, yet their value as a biomarker for aging has not been clearly established. While there have been previous reports on developing AI models to estimate age using patients' chest X-rays, these models could not be used as biomarkers for aging since they relied on images from individuals with diseases. Therefore, we aim to develop and assess the utility of a deep-learning model to estimate age using chest X-ray images from multi-institutional health checks.

Our research is divided into two stages.
The first stage involves the development and validation of an AI model that estimates age using chest X-ray images from our collected health check data as a biomarker of aging.
In the second stage, we use the chest X-ray images of patients with diseases to estimate their age using our developed AI model. We then investigate the relationship between estimated age and diseases, mortality rates, and compare with existing indicators.
Aims

The goal of the proposed study is to develop an AI model that uses chest X-ray images as a biomarker of aging and to evaluate the utility of this biomarker.

We are currently working on the first stage, developing an AI model. We used cases from multi-institutional health checks with chest X-ray images, excluding cases with registered disease history. Our AI model performed with a correlation coefficient of 95% in age estimation on the external test data set.

Aim 1: Validating the accuracy of the AI model using a multi-ethnic data set. Our primary data set mainly contains images from Japanese individuals; therefore, we aim to assess whether the accuracy is maintained when applied to a multi-ethnic dataset.

Aim 2: Verifying the relationship between estimated age and disease with logistic regression analysis using the PLCO datasets. Our external data set tests showed a relationship between estimated age and chronic diseases such as hypertension and COPD. We hypothesize that a similar correlation will be observed in the multi-ethnic external dataset.

Aim 3: Verifying the relationship between the estimated age and mortality rate of malignant diseases. Previous literature has shown a relationship between estimated age and all-cause mortality. We hypothesize that there is a correlation between the estimated age and the mortality rate of each malignant tumor.

Aim 4: We compare the predictive ability of existing prognostic indicators with the estimated age of our AI model. To our knowledge, there are no previous studies comparing existing prognostic indicators with estimated age. We hypothesize that the addition of estimated age to existing indices will be more predictive of prognosis than staging alone.
If presumed age proves to be a useful biomarker, it could be a valuable tool in future clinical practice.

Collaborators

Daiju Ueda M.D., Ph.D. Osaka Metropolitan University
Toshimasa Matsumoto Ph.D. Osaka Metropolitan University
Shannon L Walston M.S. Osaka Metropolitan University
Hiroyuki Tatekawa M.D., Ph.D. Osaka Metropolitan University
Hirotaka Takita M.D., Ph.D. Osaka Metropolitan University
Akira Yamamoto M.D., Ph.D. Osaka Metropolitan University
Yukio Miki M.D., Ph.D. Osaka Metropolitan University