Abstract
Prostate biopsy is a gold standard for diagnosing prostate cancer. In clinic, multi-needle saturation puncture is often used in the diagnosis of prostate cancer. Although it can improve the positive rate of diagnosis, it also increases the probability of postoperative infection, hematuria and other complications. This paper presents a method to identif prostate cancer by histogram of oriented gradient (HOG) and local binary pattern (LBP) feature extraction. Firstly, Gaussian filtering, gradient transformation function and other algorithms are used to preprocess the transrectal ultrasound prostate images to filter out image noise and improve contrast. Then, the local and global texture feature information of the image is extracted by using HOG and LBP. Finally, support vector machine (SVM) is used to classify features and identify positive regions. The results show that the proposed method is superior to other methods. The transrectal ultrasound prostate images exhibit superior diagnostic performance with an accuracy of 72.2% and a specificity of 75%. Experiments show that this method can provide the necessary auxiliary information for doctor diagnosis and reduce the number of puncture needles.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have