veterinary B-ultrasound Image Segmentation Region Growing Method

There are many methods for veterinary B-ultrasound image segmentation, among which the region growing method is a commonly used method. Its steps and principles are as follows.

The veterinary B-ultrasound image segmentation region growing method is a serial segmentation method. The characteristic of the serial segmentation method is to decompose the processing process into multiple steps in sequence, in which the processing of the subsequent steps is determined based on the results of the previous steps. The judgment is based on the pre-defined criteria.

The basic idea of the B-ultrasound image segmentation region growing method is to group pixels with similar properties to form a region. In the simplest form of this method, a seed is first artificially given as the starting point of growth, and then the pixels in the neighborhood around the seed pixel with the same or similar properties as the seed pixel (determined according to some predetermined growth or similarity criteria) are merged into the region where the seed pixel is located. These new pixels are treated as new seed pixels and the above process is continued until all the pixels that meet the conditions are included, so that a region is grown.

The veterinary B-ultrasound image segmentation region growing method is generally not used alone, but is placed in a series of processing processes, especially used to depict small and simple structures such as tumors and wounds. Its main drawback is that each area to be extracted must be given a seed point manually. If there are multiple areas, the corresponding number of seeds must be given. This method is also very sensitive to noise, which will cause holes or even discontinuous areas. On the contrary, local and large-scale influences will also connect the originally separated areas.

Different methods for veterinary B-ultrasound image segmentation

The classifier method is a statistical pattern recognition method used to distinguish feature spaces derived from veterinary B-ultrasound image data with known labels. Because the classifier method requires known manual segmentation results as training samples, it is a supervised pattern recognition method; the basic principle of the clustering method is roughly the same as the classifier method, the difference is that it does not require training sample data, so it is an unsupervised pattern recognition method. In order to make up for the lack of training data, the clustering method repeatedly does two things: segmenting B-ultrasound images and characterizing the characteristics of each class, so as to use the existing data to train itself to achieve the purpose of segmentation.

Artificial neural network method: The artificial neural network method uses a large number of parallel neural networks to achieve the purpose of veterinary B-ultrasound image segmentation. These networks are composed of nodes or elements that simulate biological learning mechanisms, and each node in the network can perform the most basic operations. By adjusting the weights between nodes, the network can learn biological mechanisms.

The active contour model method, also known as the variable model method, is a model-based veterinary B-ultrasound image segmentation method that uses closed parametric curves or surfaces to depict boundaries. Its original idea comes from the physical concept: in order to depict the boundary of an object, first set an initial surface or curve that is not far from the real curve or surface, and under the action of external and internal forces, push this surface or curve to move, and finally stop at the lowest energy point of the veterinary B-ultrasound image. Because the movement of the curve or surface is similar to that of a snake, this model is also called the Snake model. The active contour model has been proven to be very successful for the segmentation of veterinary B-ultrasound images.

link: https://www.bxlimage.com/nw/736.html