Automatic Cardiac Disease Assessment on cine-MRI via Time-Series Segmentation and Domain Specific Features

Abstract

Cardiac magnetic resonance imaging improves on diagnosis of cardiovascular diseases by providing images at high spatiotemporal resolution. Manual evaluation of these time-series, however, is expensive and prone to biased and non-reproducible outcomes. In this paper, we present a method that addresses named limitations by integrating segmentation and disease classification into a fully automatic processing pipeline. We use an ensemble of UNet inspired architectures for segmentation of cardiac structures such as the left and right ventricular cavity (LVC, RVC) and the left ventricular myocardium (LVM) on each time instance of the cardiac cycle. For the classification task, information is extracted from the segmented time-series in form of comprehensive features handcrafted to reflect diagnostic clinical procedures. Based on these features we train an ensemble of heavily regularized multilayer perceptrons (MLP) and a random forest classifier to predict the pathologic target class. We evaluated our method on the ACDC dataset (4 pathology groups, 1 healthy group) and achieve dice scores of 0.945 (LVC), 0.908 (RVC) and 0.905 (LVM) in a cross-validation over the training set (100 cases) and 0.950 (LVC), 0.923 (RVC) and 0.911 (LVM) on the test set (50 cases). We report a classification accuracy of 94% on a training set cross-validation and 92% on the test set. Our results underpin the potential of machine learning methods for accurate, fast and reproducible segmentation and computer-assisted diagnosis (CAD).

An iterative multi‐path fully convolutional neural network for automatic cardiac segmentation in cine MR images

“…For the segmentation of RV, the proposed method obtains a Dice value of 0.95 which is slightly higher than the top performing method. Regarding the ES phase, the performance of our method outperforms the method of Wolterink 40 in most cases and falls behind the method of Isensee et al 20 While following the one-pixel criterion described in the study of Bernard et al, 1 which is a range of agreement of 2.26 mm for the HD and 0.02 for Dice metric, the proposed method is within the range of agreement of the top performing method for all the 12 metrics.…”

“…Furthermore, we compare the performance of our method with the two ensemble model‐based methods proposed by Wolterink et al and Isensee et al . The former used an ensemble of six identical 2D CNN models.…”

“…Furthermore, we compare the performance of our method with the two ensemble model-based methods proposed by Wolterink et al 40 and Isensee et al. 20 The former used an ensemble of six identical 2D CNN models. The latter employed an ensemble comprising of ten different models based on 2D-3D U-Net architectures and was the overall top performing method of the ACDC Challenge.…”

“…However, in the context of cardiac image segmentation, due to the low through-plane resolution characteristic of cardiac MRI and the shortcomings of 3D methods such as reduction of training images and high risk of overfitting, 15,18 the segmentation performance of 3D methods may be limited to some extent. For instance, the previous work [19][20][21] evaluated on the automatic cardiac diagnosis challenge (ACDC) dataset indicated that the proposed 3D models did not meet the expectations in performance improvement over the corresponding 2D models. These motivate us to explore how to effectively leverage spatial context in 2D methods to combine the advantages of 2D and 3D methods.…”

A distance map regularized CNN for cardiac cine MR image segmentation

“…Our 2D DMR‐UNet model is able to perform as good or better than the other two 2D methods; however, the combination of 2D and 3D context has marginal improvement in the Dice overlap metric. Based on this observation, we believe the ensemble of 2D and 3D DMR‐UNet model should be able to perform as good or better than, which is not the main objective of this work. Nonetheless, we can observe the constraint imposed by the DM regularization is successful in reducing the errors in apical/basal regions, manifested in the improved Hausdorff distance.…”