Fig. 2: Estimation accuracy of the deep learning model and human physicians.
From: Deep learning-based age estimation from chest X-rays indicates cardiovascular prognosis
Estimated age by the trained deep learning model (X-ray age) in the test dataset (n = 613) (a), JSRT dataset (n = 245) (b), and estimated age by human physicians and radiologists in the JSRT dataset (n = 245) (c). Scatter plots of the chronological age (x-axis) and estimated age (y-axis) with Pearson’s correlation coefficient and intraclass correlation coefficient case 1 (ICC) and 95% confidence interval are shown. A strong positive correlation between the chronological and estimated ages was observed in the deep learning model. For the human estimation, the estimated age is the average of the estimations by the four physicians and three radiologists. The correlation between the chronological and estimated ages is modest (c).
