Quantification of Thoracic Lymphatic Flow Patterns Using Dynamic Contrast-enhanced MR Lymphangiography.
Qiang ZhengMaxim ItkinYong FanPublished in: Radiology (2020)
Background Dynamic contrast material-enhanced MR lymphangiography has recently emerged as a technique to image the lymphatic anatomy and identify lymphatic flow abnormalities; however, a method to quantify lymphatic flow in health and disease is needed. Purpose To develop a method to quantify thoracic lymphatic flow patterns using dynamic contrast-enhanced MR lymphangiography. Materials and Methods The following patients with dynamic contrast-enhanced MR lymphangiography images collected in 2015 and 2016 were retrospectively identified: group A, neonates with chylothorax; group B, children with heart failure complicated by plastic bronchitis; and group C, adults with lymphatic plastic bronchitis and without heart failure. An automated image segmentation method was developed for segmenting the contrast-enhanced lymphatic flow in spatiotemporal domains from the dynamic contrast-enhanced MR lymphangiography images. The lymphatic flow rates were quantified for individual patients on the basis of their spatiotemporal dynamic contrast-enhanced MR lymphangiography segmentation results, and the flow rates were compared among the three patient groups by using Wilcoxon rank sum tests. Results Twenty-two patients were evaluated: seven neonates (mean age, 49 days ± 71 [standard deviation]; three boys, four girls), 10 children (mean age, 8 years ± 3; seven boys, three girls), and five adults (mean age, 46 years ± 10; three men, two women). The proposed method was used to obtain lymphatic flow segmentation results with Dice scores of 0.80, 0.82, and 0.83 for patients from groups A, B, and C, respectively. The mean flow rates for groups A, B, and C were 1.8 mL/min ± 1.4, 4.0 mL/min ± 1.8, and 12.5 mL/min ± 3.8, respectively. The flow rate differed significantly between groups A and B (P = .002), groups A and C (P = .01), and groups B and C (P = .01). Conclusion An automatic spatiotemporal segmentation method was used to determine thoracic lymphatic flow rates in individual patients based on their dynamic contrast-enhanced MR lymphangiographic images. © RSNA, 2020 Online supplemental material is available for this article.
Keyphrases
- contrast enhanced
- deep learning
- lymph node
- end stage renal disease
- magnetic resonance
- heart failure
- convolutional neural network
- newly diagnosed
- chronic kidney disease
- magnetic resonance imaging
- computed tomography
- peritoneal dialysis
- public health
- prognostic factors
- type diabetes
- spinal cord
- machine learning
- pregnant women
- climate change
- left ventricular
- skeletal muscle
- risk assessment
- case report
- optical coherence tomography
- insulin resistance
- low birth weight
- preterm birth