Diagnosing thyroid cancer with ultrasound. Like other organs, the thyroid gland can develop growths, or nodules, that can turn malignant. Thanks to the gland’s accessibility, thyroid nodules are readily detected using ultrasound. Unfortunately, to determine whether a nodule is malignant, a doctor must insert a needle into the patient’s neck and draw out a sample of cells. Not only is the procedure costly (about $1500 in the US), it’s more likely than not to confirm the general finding that most nodules—around 70%—are benign. Malignant nodules are stiffer than either benign nodules or thyroid tissue. Being a mechanical stimulus, ultrasound can sense differences in elasticity. But can it diagnose the malignancy of nodules as well as detect them? According to Yongmin Kim (University of Washington in Seattle and POSTECH in South Korea) and his collaborators, the answer is yes. Kim’s team has devised an algorithm that optimizes the use of routinely gathered ultrasound data. The algorithm calculates two metrics: the pixel-by-pixel absolute strain rate and a pixel-to-pixel measure of local contrast. From the metrics, the algorithm creates an elasticity contrast map. The incorporation of local contrast is especially helpful in revealing small malignant nodules whose telltale stiffness can otherwise be masked by the greater elasticity of surrounding tissue. In tests on patients whose thyroid nodules had been independently assessed for malignancy, the algorithm performed well: 19 out of 20 malignant nodules and 76 out of 103 benign nodules were correctly identified. (S. Luo, D.-J. Lim, Y. Kim, Med. Phys. 39, 1182, 2012.)
This article does not have any references.