Yao, Tina;
St. Clair, Nicole;
Miller, Gabriel F;
Dorfman, Adam L;
Fogel, Mark A;
Ghelani, Sunil;
Krishnamurthy, Rajesh;
... Muthurangu, Vivek; + view all
(2023)
A Deep Learning Pipeline for Assessing Ventricular Volumes from a Cardiac Magnetic Resonance Image Registry of Single Ventricle Patients.
Radiology: Artificial Intelligence
10.1148/ryai.230132.
(In press).
Preview |
Text
yao-et-al-2023-a-deep-learning-pipeline-for-assessing-ventricular-volumes-from-a-cardiac-magnetic-resonance-image.pdf - Accepted Version Download (5MB) | Preview |
Abstract
Purpose: To develop an end-to-end deep learning (DL) pipeline for automated ventricular segmentation of cardiac MRI data from a multicenter registry of patients with Fontan circulation (FORCE). / Materials and Methods: This retrospective study used 250 cardiac MRI examinations (November 2007–December 2022) from 13 institutions for training, validation, and testing. The pipeline contained three DL models: a classifier to identify short-axis cine stacks and two UNet 3+ models for image cropping and segmentation. The automated segmentations were evaluated on the test set (n = 50) using the Dice score. Volumetric and functional metrics derived from DL and ground truth manual segmentations were compared using Bland-Altman and intraclass correlation analysis. The pipeline was further qualitatively evaluated on 475 unseen examinations. / Results: There were acceptable limits of agreement (LOA) and minimal biases between the ground truth and DL end-diastolic volume (EDV) (Bias: -0.6 mL/m2, LOA: -20.6–19.5 mL/m2), and end-systolic volume (ESV) (Bias: - 1.1 mL/m2, LOA: -18.1–15.9 mL/m2), with high intraclass correlation coefficients (ICC > 0.97) and Dice scores (EDV, 0.91 and ESV, 0.86). There was moderate agreement for ventricular mass (Bias: -1.9 g/m2, LOA: -17.3–13.5 g/m2) and a ICC (0.94). There was also acceptable agreement for stroke volume (Bias:0.6 mL/m2, LOA: -17.2–18.3 mL/m2) and ejection fraction (Bias:0.6%, LOA: -12.2%–13.4%), with high ICCs (> 0.81). The pipeline achieved satisfactory segmentation in 68% of the 475 unseen examinations, while 26% needed minor adjustments, 5% needed major adjustments, and in 0.4%, the cropping model failed. / Conclusion: The DL pipeline can provide fast standardized segmentation for patients with single ventricle physiology across multiple centers. This pipeline can be applied to all cardiac MRI examinations in the FORCE registry.
Archive Staff Only
 |
View Item |
