Segmentation of Thyroid Nodules Based on Hyperspectral Images

Thyroid nodules are a common thyroid disease, and early detection and treatment have a significant impact on the prognosis of patients. Hyperspectral image technology has significant advantages in medical image analysis due to its high-resolution and multi band characteristics, especially in the recognition and classification of thyroid nodules, which has important application value. But currently, there is no very effective method for segmenting hyperspectral images of thyroid nodules. Deep learning algorithms have shown superior performance in recent years in computer vision and pattern recognition tasks, including medical image analysis. This paper proposes the use of VGG model for segmentation of hyperspectral images of thyroid nodule tissue. Hyperspectral images can obtain spatial and spectral information of organizations, and VGG models can segment images through deep networks. Using hyperspectral and VGG, we found that thyroid nodule tissue can be well distinguished at the pixel level.