Thyroid Nodules Image Fusion Based On Shearlet Transform And Sparse Representation

Thyroid nodule is a kind of common thyroid disease with high incidence, but only about five percent of the thyroid nodule is malignant, timely detection and treatment can be cured completely usually. So, correct detected the benign or malignant of thyroid nodule is of great significance. B-type Ultrasound examination becomes the preferred method of thyroid disease inspection, due to its no-radiation and high detection rate, which can provide the morphological changes in thyroid diseased tissue, but it cannot judge the function. SPECT image is a kind of functional image which can provide the function changes in diseased tissue.However, SPECT imaging cannot locate the thyroid nodes with accuracy. The image fusion technique can be used to fusing the two kinds of image, which help doctors understand complementary information of the diseased tissue and diagnose the thyroid nodule comprehensively.This dissertation takes the B-type Ultrasound image and SPECT image as the investigative object, start with two aspects--multi-scale geometric analysis tool and fusion rule, an image fusion algorithm based on Shearlet transform and sparse representation is proposed. Shearlet transform can overcome the shortcoming of wavelet transform which restrictions on the number of directions and can capture the geometry information of multidimensional data efficiently. Nevertheless, the sparsity of low frequency sub-band is inferior. So, the sparse representation theory is introduced in fusion process under the foundation of the Shearlet transform.The sparse representation is introduced in low frequency fusion process, then the rule of selecting the fused frequency sparse coefficients that the energy is big in order to keep the details as much as possible. The fused high frequency sub-band coefficients based on region sum modified laplacian which utilizing the relevant of pixels. Make some fusion experiments for B-type Ultrasound image and SPECT image of thyroid nodule, the results show that the proposed algorithm has better performance than the image fusion algorithm of the multi-scale and the sparse representation in single scale, both have the ability of sparse representation data and analysis data in multi-scale.