Cascaded coarse-to-fine convolutional neural networks for pericardial effusion localization and segmentation on CT scans

Abstract

Pericardial effusions can impair cardiac function so detecting them is important. On CT scans, pericardial effusions demonstrate very high shape and volume variability and very low contrast to adjacent structures. This inhibits traditional automated segmentation methods from achieving high accuracies. Deep neural networks have been widely used for image segmentation in CT scans. In this work, we present a two-stage method for pericardial effusion localization and segmentation. For the first step, we localize the pericardial area from the entire CT volume, providing a reliable bounding box for the more refined segmentation. A coarse-scaled convolutional networks model is trained on entire CT volume. The resulting per-pixel probability maps are then thresholded to produce a bounding box covering the pericardial area. For the second step, a fine-scaled convolutional networks model is trained only on the bounding box region for effusion segmentation to reduce the background distraction. Two neural network architectures, the holistically-nested convolutional networks (HNN) and U-Net were compared. Quantitative evaluation is performed on a dataset of 25 CT scans (1206 images) of patients with pericardial effusions. The segmentation accuracies of our two-stage method, measured by Dice Similarity Coefficient, were 75.59±12.04% for HNN and 77.33±10.02% for U-Net. Both were significantly better than the segmentation accuracy (62.74±15.20% for HNN and 74.81±11.01%) of only using the coarse-scaled model.