Deep Learning Models for Calculation of Cardiothoracic Ratio from Chest Radiographs for Assisted Diagnosis of Cardiomegaly

Abstract

We propose an automated method based on deep learning to compute the cardiothoracic ratio and detect the presence of cardiomegaly from chest radiographs. We develop two separate models to demarcate the heart and chest regions in an X-ray image using bounding boxes and use their outputs to calculate the cardiothoracic ratio. We obtain a sensitivity of 0.96 at a specificity of 0.81 with a mean absolute error of 0.0209 on a held-out test dataset and a sensitivity of 0.84 at a specificity of 0.97 with a mean absolute error of 0.018 on an independent dataset from a different hospital. We also compare three different segmentation model architectures for the proposed method and observe that Attention U-Net yields better results than SE-Resnext U-Net and EfficientNet U-Net. By providing a numeric measurement of the cardiothoracic ratio, we hope to mitigate human subjectivity arising out of visual assessment in the detection of cardiomegaly. Related WorkComputer aided-diagnosis has been used to detect cardiomegaly in a chest X-ray with or without using deep learn-

Cardiothoracic ratio values and trajectories are associated with risk of requiring dialysis and mortality in chronic kidney disease

“…Furthermore, despite ongoing efforts to investigate the prognostic role of the CTR in various disease states, the validity and generalizability of these studies are uncertain because sample sizes are often small; CTR measurement requires labor-intensive annotation by physicians. Although several machine learning (ML) algorithms have been developed to simplify and standardize the annotation process, adoption of these algorithms has been slow because the algorithms require further validation before they can be fully integrated into the clinical workflow [16][17][18][19][20] . To address this research gap and evaluate the risks of requiring dialysis and mortality associated with the various baseline CTRs and CTR trajectories in patients with CKD-ND, we developed an artificially intelligent CTR (iCTR) assessment system to annotate PA-CXRs.…”