Research And Application Of Ultrasound Image Recognition Method Based On Phase Information

With the development of ultrasound imaging technology,medical ultrasound image automatic recognition algorithms have been developed rapidly.The image quality problems caused by the limitations of ultrasound imaging equipment and the lack of operator's experience and other reasons make the research of computer-aided diagnosis of great significance,so the research of this paper is launched.The image processing algorithm based on phase information is insensitive to illumination change and the local contrast,so it can recognize image features stably.In this paper,the principle of phase congruence and phase symmetry algorithm is deeply studied and applied to ultrasound image processing.The main research contents are as follows:(1)Based on the phase symmetry(PS)algorithm,this paper proposes a monogenic phase symmetry-and confidence map-based method for recognition of ultrasonic distal radius bones.First,the phase symmetry algorithm is applied to recognize the image features,so as to detect the edge and remove the noise.At the same time,the confidence map of the shadow behind the bone in the ultrasound image is obtained by using the confidence map model to confirm the position of the bone surface.Finally,the bone surface structure is obtained by multiplying the phase symmetry feature map and the confidence map.The experimental results of radial ultrasound images show that this method can successfully identify the bone surface.(2)In this paper,an algorithm based on phase symmetry feature is proposed to segment the diaphragm.An automatic method to calculate the diaphragmatic excursion(DE)is further presented.In order to verify the effectiveness of the proposed method,twenty diaphragmatic ultrasound cines are employed in this study.Both the segmentation and DE results are compared with the gold standard.Results demonstrate that the proposed technique segments the diaphragm with a localization mean error of 0.73 mm.Furthermore,a coefficient of determination(R-squared)of 0.832 is obtained between the automatic and manual measurement of DE.(3)Based on the PS and Hough transform,this paper proposes a needle location and trajectory prediction algorithm.First,the PS algorithm is used to identify the edge structure in the ultrasound image.After that,the edge image is obtained by Hough transform.Finally,the path of the needle is predicted by the identified direction of the needle.The experiment of phantom biopsy proves that the algorithm proposed in this paper can accurately identify the biopsy needle.(4)Based on the original phase symmetry algorithm,multi-degree phase symmetry(MDPS)model,which is a modified phase symmetry algorithm,is proposed to enhance microcalcification in ultrasound breast images.First,Log-Gabor filter is applied to construct multi-degree and multi-scale symmetric filters.Applying these filters to images,multi-scale and multi-degree outputs are obtained.Adding together these multi-scale outputs and multiplying these multi-degree outputs to select the microcalcification candidate points.Then through threshold process,the final microcalcification points are gained.In the end,the identified microcalcification is enhanced on the original image.