Flow Quantification from 2D PC-MRI in Renal Arteries using Clustering

F.G. Zöllner^1,2, J. A. Monssen³, J. Roervik^2,3, A. Lundervold^4
1Computer Assisted Clinical Medicine, Faculty of Medicine Mannheim, University of Heidelberg, Germany
²Dept. of Surgical Sciences, University of Bergen, Norway
³Dept. of Radiology, Haukeland University Hospital , Norway
⁴Dept. of Biomedicine, University of Bergen,Norway

Introduction

• Renal Disease caused e.g. by hypertension
• Renal Artery Stenosis (RAS) leading cause for renal hypertension
• Loss of renal parenchyma in the stenosed kidney
• Microangiopathy in glomeruli in non-stenosed contralateral kidney
• Important parameter for choosing patients to treat by PTA:
  • Kidney volume
  • Flow reduction
  • Flow velocity
  • Renal function
• Cine Phase Contrast MRI tool for grading stenosis
• Can be used for assessing blood flow and velocity

Methods

• Today: mainly manual ROI drawing in magnitude images, frame-by-frame
• Automated procedures:
  • Active contours
  • Correlation and manual thresholding
• K-Means clustering of velocity profiles in phase images
• Random initialisation of algorithm, 10 repetitions
• Explored cosine distance function, Eucleadian and correlation distance

Data Acquisition

• 5 subjects (3 healthy volunteers, 2 patients)
• ECG gated 2D Cine PC-MR sequence
• TE=4ms,TR=37ms, FA=30°, VENC=100 cm/s
• 20-25 images per cardiac cycle, RR-Interval 758-1072ms, 2 averages, matrix 256x192
• Slice thickness 6mm, in-plane resolution <0.5mm2
• Transec. slice through vessel 1-2cm distal of stenosis

Results

• Descrimination of vessel from background for K=2
• Analysis of optimal cluster number:
  • 2 local minima for DB-Index
  • K=3-5 and K=8-10 depending on the subject
• For K>2 vessel area devided into subcluster
• Cosine distance function yielded best results
• Phase wrap included in one dataset showed up as a seperate cluster
• Comparison manual vs. clustering showed high similarity in the velocity profiles

Discussion

• Semi-automatic approach using K-Means clustering
• Reduced manual interaction
• Only number of clusters has to be given
• Obtained velocity profiles similar to manual delineations
• Detection of different blood flow patterns at higher K values, i.e. more allowed clusters
• Interpretation needs further investigation
• Conclusion: approach can aid in clinical assessment and grading of renal artery stenosis