A dermatoscope image segmentation method based on a multi-branch convolutional neural network

Abstract

The invention discloses a dermatoscope image segmentation method based on a multi-branch convolutional neural network. The method comprises the following steps of 1, collecting training samples; 2, expanding the image; 3, designing a multi-branch convolutional neural network model; 4, training the multi-branch convolutional network; 5, generating a skin damage distribution probability graph; 6, obtaining a segmentation result. The method has the advantages that the training data set is effectively expanded by using the corresponding image transformation according to the data characteristics ofthe dermatoscope image, so that the network training is effective, and the generalization performance is high; the convolutional neural network comprises a plurality of branches, rich semantic information and detail information are fused, compared with a common network, the skin lesion edge can be better recovered, and a more accurate skin lesion segmentation result is obtained; the method is a full-automatic segmentation scheme, only the dermatoscope image to be segmented needs to be input, the segmentation result of the image can be automatically given through the scheme, the additional processing is not needed, and the method is efficient, simple and convenient.

Description

technical field [0001] The invention belongs to the field of computer-aided diagnosis, and in particular relates to a method for segmenting dermoscopic images based on a multi-branch convolutional neural network. Background technique [0002] Skin melanoma is divided into benign and malignant, among which malignant skin melanoma is extremely harmful. If patients do not receive timely treatment in the early stage, it will easily lead to death. For skin melanoma, the most effective treatment is to detect it as early as possible and then perform lesion resection. Dermoscopy, also known as epidermal light transmission microscope, can obtain dermoscopic images with high resolution and clarity. Automatic diagnosis of dermoscopic images can avoid diagnostic errors caused by subjectivity in doctors' diagnosis. When diagnosing skin lesions, regional shape and boundary information are important diagnostic basis. The purpose of dermoscopy image segmentation is to obtain accurate les...

AI Technical Summary

Problems solved by technology

The method based on supervised learning manually designs relevant features, or automatically mines potential features from data, and then trains a classifier to classify features, and judges whether a sub-region or a certain pixel belongs to skin lesions or normal skin; this method is very dependent on feature design and selection. , Poor adaptability to complex dermoscopic images

Images