# Thyroid nodules classification in ultrasound imaging using artificial intelligence methods _by Xia Lumi (IETR - Vaader) - 2022.02.24_ ###### tags: `VAADER` `Seminar` ![](https://i.imgur.com/qMEYnl7.jpg) ## Abstract The precise diagnosis of thyroid nodules remains a tricky problem despite improvements of medical technologies. With the aim to improve the diagnostic efficiency of thyroid nodules, we explored here the use of artificial intelligence methods to classify thyroid nodules using ultrasound images and related structured data. With a database including exams of about 100 patients, imaged with 3 ultrasound modalities (B-Mode, Doppler, Elastography) as well as structured data (personal information, dimensions of nodules, elasticity measurements, SUV measurements etc), we established an artificial intelligence scheme for classifying the nodules. We measured the importance of different modality images in the classification and found that SUV measurements and dimension of nodules are the most influential factors among the structured data. In general, Doppler images and Elastography images showed a better classification performance than B-Mode images. Elasticity shows a correlation with the malignancy of thyroid nodules from both mathematical study and our classification scheme.