| With the continuous development of ultrasound medical technology,ultrasound therapy plays an increasingly extremely important role in the clinical medical service system due to its non-invasive,real-time and low cost.In order to obtain good ultrasound treatment results,it is particularly important to accurately characterize the texture features of ultrasound images.However,due to the influence of the ultrasound imaging system,etc.,the ultrasound image contains a texture that is perpendicular to the ultrasound beam and whose direction changes with position,we call this directional pseudo texture.The presence of directional pseudo-texture reduces the resolution of the ultrasound image,and at the same time makes the extracted ultrasound image features contain interference direction information,which affects the correctness of subsequent diagnosis.In this paper,a method based on directional pseudo-texture correction for ultrasonic image feature extraction is proposed,the method firstly splits the ultrasound image with speckle noise removed by filtering to generate super-pixels,and obtains the same super-pixel block on the ultrasound image by mark,then performing a Fourier transform on the minimum circumscribed rectangular pixel block of super-pixel complemented by each pixel,and performing ellipse fitting on the transformed spectrum to obtain a deflection angle of the corresponding spectrum,thereby obtaining a texture direction of the super-pixel,rotating the corresponding super pixel block to the same direction according to the deflection angle,and the texture features of the ultrasonic image are extracted by combining the feature of the gray level co-occurrence matrix.In order to demonstrate the effectiveness of the ultrasound image feature extraction method based on directional pseudo-texture correction proposed in this paper,we validated this on a tumor segmentation framework based on splitting and merging,the directional texture correction ultrasonic image feature extraction method and the directional pseudo texture uncorrected ultrasound image feature extraction method are applied to the tumor segmentation process,respectively.Then using the results of tumor segmentation and evaluation indicators to qualitatively and quantitatively analyze the proposed algorithm,and prove the effectiveness of the proposed algorithm from the side.The algorithm proposed in this paper has carried out experiments on the ultrasound-guided uterine fibroids image based on the HIFU-treated,the results show that no matter the number of the ultrasound images from which the tumor can be segmented or the accuracy of the tumor segmentation result,the texture features based on directional pseudo-texture correction are superior to the directional pseudo-texture uncorrected texture features,Therefore,the method of ultrasonic image feature extraction based on directional pseudo-texture correction proposed in this paper is reliable and effective,and it is promising to be applied in the process of ultrasonic image feature extraction for medical treatment. |
