A research group led by Yohito Mitsuyama, a second-year doctoral student at the Graduate School of Osaka Public University, developed an AI model that estimates the body age from chest X-ray images. demonstrated a relationship between different diseases.
Due to the complexity of age-related changes, various biomarkers of aging have been proposed for understanding aging.Although there have been reports of estimating age with AI using chest X-ray images, his AI model based on chest X-ray images collected from multiple facilities has not been developed.
Therefore, the research group developed and trained an AI model for age estimation, and externally tested 3 chest X-ray images of 67,099 healthy subjects (those with a medical history were excluded) collected from three facilities. used.As a result, the AI model showed very high estimation accuracy (correlation coefficient of 36,051, and 0.95 is usually regarded as a very strong correlation).In addition, we suggested the possibility that there is a basis for judging aging in the lower lung field and aortic arch from the visualized images when AI estimates age.
Furthermore, this AI model was used to analyze the disease and age difference by odds ratio using chest radiographs of 2 patients with the disease collected from two other institutions.We found that the higher the estimated age than the actual age, the more positively correlated with the incidence of chronic diseases such as hypertension, hyperuricemia, chronic obstructive pulmonary disease, and interstitial pneumonia.
This study suggests that age expressed on chest radiographs may more accurately reflect health information than biological age.In the future, we will proceed with the development of AI biomarkers for estimating the severity of chronic diseases, predicting life expectancy, stratifying the prognosis of malignant tumors, and predicting surgical complications.
