Segmentation and Skeletonization of 3D Contrast Enhanced Ultrasound Images for the Characterization of Single Thyroid Nodule

The thyroid nodules are common findings in clinical practice. Recent studies demonstrated that thyroid nodules can be found in about 66% of the adult population. Nevertheless, only 5–7% of the thyroid nodules is malignant. In the presence of a suspicious nodule, thyroidectomy is performed. As a consequence, the patient will lack the thyroid hormones, which will be integrated by using drugs. Currently, the differential diagnosis of thyroid nodules is still problematic. The most used technique is the ultrasound examination, but it has been proved that several interpretative limitations of the ultrasound nodule appearance still remain. Color and Power Doppler slightly improved the accuracy of ultrasound characterization, but performance is still inadequate for clinical practice. Other possible diagnostic techniques are MRI and nuclear medicine. MRI is expensive and still not very diffuse for thyroid analysis, while nuclear medicine essentially differentiates cold from hot nodules. In this chapter, we will describe a novel strategy for the characterization of thyroid nodules based on 3-D ultrasound images acquired after the injection of intravascular contrast agent. The objective of this processing technique is the characterization of the intranodular vascularization, under the hypotheses that malignant nodules had a greater vascularization than benign. The processing technique consists of three steps: (1) preprocessing and segmentation of the contrast agent distribution in the nodule volume; (2) thinning (by means of a combination of Distance Transform and Ma and Sonka skeleton) of the segmented 3-D volumes; and (3) computation of intranodular vascularization descriptors. Our results showed that malignancy is associated to a higher vascularization. Malignant nodules had a higher number of vascular trees, which had several branches and a marked tortuosity. Conversely, benign lesions were overall poorly vascularized. This segmentation and skeletonization technique is a first step toward a 3-D contrast agent-based ultrasound technique for the differential diagnosis of thyroid nodules.