According to the University of Tokyo Hospital, the research group collected data from more than 10 chest X-ray image databases in the United States and developed a model that estimates a patient's age from a single image by having artificial intelligence learn from this data.

　Using 245 pieces of data held by the Japanese Society of Radiological Technology, artificial intelligence was used to estimate the patient's age, showing a mean absolute error of 4.95 years.Seven cardiologists, pulmonologists, and radiologists used the same data to estimate the patient's age, and found that the mean absolute error was 7 years, indicating that the artificial intelligence system outperformed human doctors.

　Furthermore, when applying this artificial intelligence model to chest X-ray images in the United States, it became clear that patients with abnormal findings such as pleural effusion and fibrosis tended to be judged older than their actual age.Data from 1,562 patients admitted to Sakakibara Memorial Hospital for heart failure also showed that patients with hypertension, atrial fibrillation, and atrial flutter were estimated to be about one year older than their actual age.

　The research group has shown the possibility of verifying the risk of patients who have been judged to have no abnormalities in the examination, and also expects that it can be applied in the medical field as a new indicator for heart disease.

Paper information:[Communications Medicine] Deep learning-based age estimation from chest X-rays indicates cardiovascular prognosis