3D Ultrasound Reconstructions of the Carotid Artery and Thyroid Gland Using Artificial-Intelligence-Based Automatic Segmentation—Qualitative and Quantitative Evaluation of the Segmentation Results via Comparison with CT Angiography

Figures

• 

  Figure 1. Manual and semiautomatic segmentation using our own software application: original US image, thyroid—yellow, carotid arterial wall—white, calcified plaque—light blue, soft plaque—dark blue, and circulated lumen—red.

• 

  Figure 2. Segmented blood vessels from CT angiography.

• 

  Figure 3. Distribution of number of masks/classes in datasets. Datasets from carotid-artery-disease patients—orange; healthy patients—blue.

• 

  Figure 4. (a) Original US image and (b) CNN-based automatic segmentation. Arterial wall—white, calcified plaque—light blue, soft plaque—dark blue, circulated lumen—red. (c) Masks generated by the operator and (d) 3D US reconstructions of the scanned area, based on masks generated with CNN-based automatic segmentation.

• 

  Figure 5. (a) The 3D US reconstruction of the carotid artery circulated lumen, (a′) the centroid of the cross-section of the circulated lumen, and (a″) representation in the 3D space of the centroids of the 2D ultrasonographic section through the carotid artery. (b) The CT angiography for the carotid artery of the circulated lumen, (b′) the transversal section through the carotid arteries’ CT scans, and (b″) extracted centroids of the carotid arteries, from the CT angiography represented in the 3D space.

• 

  Figure 6. Patient 1, scan 2, circulated lumen evaluated with Patient 1, scan 1, circulated lumen as reference, aligned using CCOSS. The 100% overlapped 3D points are blue; the most distant points are represented in red.