The thyroid is a hormone-producing gland located at the front of the neck, below the larynx. Diseases of the thyroid gland can be divided into two main groups: those that involve a change in hormone production and hormone secretion (hyperthyroidism and hypothyroidism), and those that affect the shape and size of the gland (goiter, thyroid cancer) [1]
In ultrasound (US) medical imaging, thyroid cancer manifests itself as irregular tissues, called nodules. There are however different types of nodules that look similar but are caused for different reasons than malignant cancer. It is therefore crucial to correctly classify these nodules to select a proper treatment.
When the nodule appearance in the US image is not obvious or distinguishable, typically an ultrasound guided fine needle biopsy with cytological evaluation (FNC) is needed to diagnose the nodule. A study performed by UNN radiologists [2] concludes that dedicated expert-analysed US imaging without FNC can reliably distinguish benign versus malignant nodules, and also differentiate between several histopathological entities in thyroid nodules. Because there are enough relevant visual features in US images, there is potential for a reduction in the number of invasive and time-consuming FNC biopsies and diagnostic operations.
Artificial intelligence provides tools that can help us to correctly classify nodules and estimate the probability of malignant cancer. Therefore, SPKI is starting up this project in collaboration with UNN, Helse Nord IKT and Visual Intelligence (UiT) to develop a US decision support system for patients at Helse Nord.
