Breast tumor feature selection method based on Relief algorithm

Abstract

The invention relates to a breast tumor feature selection method based on a Relief algorithm. The method comprises two parts: a breast tumor feature extraction module and a Relief feature selection module. The breast tumor feature extraction module is used for extracting breast tumor features, extracting numerical value capable of describing breast tumor features from a breast tumor image, and dividing the tumor features into three major classes, that is, morphological feature, textural feature and wavelet feature; and the Relief feature selection module is used for carrying out feature selection to obtain a good feature subset, thereby improving classification performance and breast tumor diagnosis accuracy; for each textural feature and wavelet feature, feature selection is carried out through the Relief feature selection algorithm, and then, splicing and fusing with all morphological features are carried out to form a plurality of feature subspaces; and the obtained feature subset is subjected to splicing and fusing with all morphological features, thereby finishing the whole breast tumor feature selection process.

Description

technical field [0001] The invention relates to data mining technology and biomedical engineering technology, in particular to a feature selection method for breast tumors. Background technique [0002] Breast cancer is the most common malignant tumor in women. Worldwide, its fatality rate ranks first among female malignant tumors. At present, early diagnosis and timely treatment are the most effective measures to deal with breast cancer. Medical imaging methods, such as X-rays, nuclear magnetic resonance, and ultrasound detection, are currently the most important means of detecting and diagnosing breast cancer. However, a large amount of image information generated in mammary gland examination is easy to make doctors tired, and the diagnostic accuracy is affected by the professional ability, experience and subjective factors of doctors. In this context, the use of machine learning methods to determine whether a tumor exists and whether it is benign or malignant has becom...

AI Technical Summary

Problems solved by technology

Complete search searches every feature in the feature set. Although it can avoid falling into local optimum, the computational complexity is extremely high, and the effect is not ideal in practical applications.

The basis of heuristic search is the greedy algorithm. Although it cannot guarantee the global optimality of the obtained feature subset, it has low computational complexity and high execution efficiency, so it has been widely used.

Images