Automatic Cardiothoracic Ratio Calculation With Deep Learning

Abstract

Deep learning is a growing trend in medical image analysis. There are limited data of deep learning techniques applied in Chest X-rays. This paper proposed a deep learning algorithm for cardiothoracic ratio (CTR) calculation in chest X-rays. A fully convolutional neural network was employed to segment chest X-ray images and calculate CTR. CTR values derived from the deep learning model were compared with the reference standard using Bland-Altman analysis and linear correlation graphs, and intra-class correlation (ICC) analyses. Diagnostic performance of the model for the detection of heart enlargement was assessed and compared with other deep learning methods and radiologists. CTR values derived from the deep learning method showed excellent agreement with the reference standard, with mean difference 0.0004 ± 0.0133, 95% limits of agreement -0.0256 to 0.0264. Correlation coefficient between deep learning and reference standard was 0.965 (P <; 0.001), and ICC coefficient was 0.982 (95% CI 0.978-0.985) (P <; 0.001). Measurement time by deep learning was significantly less than that of the manual method [0.69 (0.69-0.70) VS 25.26 (23.49-27.44) seconds, P <; 0.001]. Diagnostic accuracy, specificity, and positive predictive value were comparable between the two methods. However, deep learning showed relatively higher sensitivity and negative predictive value (97.2% vs 91.4%, P = 0.004; and 96.0% vs 89.0%, P = 0.006; respectively) compared with the manual method. Performance of this computer-aided technique was demonstrated to be more reliable, time and labor saving than that of the manual method in CTR calculation.