Arterial diameter, diffusion distance, and proportion of wrapping were manually measured for vessel pairs that were matched in both the light micrographs and micro-CT images in a blinded manner. Artery-vein pairs were identified, matched in both micro-CT and light microscopy images, and then labelled. Measurements were first made for the artery-vein pairs in the light micrographs. Then, measurements were made for the artery-vein pairs in the micro-CT images. A total of 25 vessel pairs were identified across two kidney volumes.
9.3.10 Automated analysis of the radial geometry of artery-vein pairs
Sub-volumes of the three-dimensional whole-kidney reconstructions were generated, that included large arteries (> 100 μm) that were expected to be …show more content…
4.5.2, The Mathworks Inc., Natick, MA, USA). All voxels were then subjected to a threshold to distinguish the vascular voxels from the non-vascular (background) voxels (Fig 9.3A). A multi-seed region growing process was performed for segmentation of arterial and venous voxels. At least one seed voxel was manually identified in each vascular region (arterial and venous; Fig 9.3B). Once identified, each region was designated a vessel type (artery or vein). Geodesic distances of vascular voxels from the nearest seed voxel were then generated. Each vascular voxel was then assigned a vessel type. The vessel type was the type of its nearest seed voxel. A volumetric atlas was then produced to map the vascular voxels to either ‘artery’ or ‘vein’. The atlas was then smoothed via an averaging filter. A three-dimensional vascular geometry was produced using isosurface generation. The isosurface geometry was generated as a triangulated mesh. The geometry was simplified but still preserved vascular shape. A centreline (skeleton) was extracted from the 3D geometry using mesh contraction (2) (Fig 9.3C) before automated analyses were generated (Fig
9.3.10 Automated analysis of the radial geometry of artery-vein pairs
Sub-volumes of the three-dimensional whole-kidney reconstructions were generated, that included large arteries (> 100 μm) that were expected to be …show more content…
4.5.2, The Mathworks Inc., Natick, MA, USA). All voxels were then subjected to a threshold to distinguish the vascular voxels from the non-vascular (background) voxels (Fig 9.3A). A multi-seed region growing process was performed for segmentation of arterial and venous voxels. At least one seed voxel was manually identified in each vascular region (arterial and venous; Fig 9.3B). Once identified, each region was designated a vessel type (artery or vein). Geodesic distances of vascular voxels from the nearest seed voxel were then generated. Each vascular voxel was then assigned a vessel type. The vessel type was the type of its nearest seed voxel. A volumetric atlas was then produced to map the vascular voxels to either ‘artery’ or ‘vein’. The atlas was then smoothed via an averaging filter. A three-dimensional vascular geometry was produced using isosurface generation. The isosurface geometry was generated as a triangulated mesh. The geometry was simplified but still preserved vascular shape. A centreline (skeleton) was extracted from the 3D geometry using mesh contraction (2) (Fig 9.3C) before automated analyses were generated (Fig