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

TL;DR

This paper introduces a deep learning-based automated method to measure the cardiothoracic ratio from chest X-rays, aiding in the objective detection of cardiomegaly with high accuracy.

Contribution

It presents a novel automated approach using deep learning models to accurately compute the cardiothoracic ratio, including a comparison of different segmentation architectures.

Findings

Achieved 0.96 sensitivity at 0.81 specificity on test data.

Attention U-Net outperformed other segmentation models.

Provided a reliable numeric measurement to reduce subjective assessment.

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.