Breast tumor ultrasonic image classification method based on three-dimensional convolutional neural network

Abstract

The invention discloses a breast tumor ultrasonic image classification method based on a three-dimensional convolutional neural network, and relates to the technical field of computer image processing. The method comprises the specific steps that data preprocessing is carried out, a training set is used for training a proposed three-dimensional convolutional neural network, benign tumors are marked as 0, and malignant tumors are marked as 1; testing the trained model by using a test set, and finally outputting the probability of benign and malignant tumors through a softmax function, extracting regions of interest, and performing tumor probability estimation on each region of interest by using a three-dimensional convolutional neural network; diversity, namely complexity, of breast cancerimages causes certain difficulties to diagnosis of doctors, the misdiagnosis rate of tumors can be effectively reduced through automatic classification of breast tumor ultrasonic images based on deeplearning, and meanwhile reliable reference bases can be provided for diagnosis of the doctors.

Description

technical field [0001] The invention belongs to the technical field of computer image processing, and relates to a breast tumor ultrasound image classification method based on a three-dimensional convolutional neural network, specifically using a three-dimensional convolutional neural network method to classify benign and malignant breast tumors in breast ultrasound images . Background technique [0002] Early detection and treatment of breast cancer can reduce its mortality rate. In this process, mammography, ultrasound, magnetic resonance and other images play a very important role in the diagnosis of breast cancer. Factors such as the technical characteristics of different images, image quality, and subjective judgment of doctors will affect the accuracy of breast cancer diagnosis. Therefore, it is particularly important to improve the ability and accuracy of doctors to detect lesions through computer assistance. [0003] At the same time, in recent years, the computing ...

AI Technical Summary

Problems solved by technology

Medical imaging has the defects of small amount of data, inconspicuous lesion area and large noise, and the existing convolutional neural network models are all proposed based on large-scale two-dimensional image data, which cannot adapt to medical imaging data.

Images