The WHO (World Health Organization) estimates that, in 2012, it will emerge 1.7 million new cases of breast cancer in world. Many studies aim to distinguish malignant cancers from benign. The goal of the proposed work is give to health professional more subsidies in order to analyze the patient situation, through the tumor contour classification. The lesion contour is a predominant factor in order to choose the appropriate treatment for the patient and detecting the degree of malignancy of the cancer. The proposed work classifies the lesion according the American College of Radiology rules. It is employed two groups of Zernike Moments in order to descript the tumor contour and applied to ELM and SVM Neural Networks. Different from the ELM and SVM in literature, the proposed work extends these two neural networks to kernel learning. The best result is about 80% of hit rate, using SVM with a RBF kernel.
30 October 2020 in Progress in Artificial Intelligence
According to the World Health Organization, breast cancer is the most common type of cancer in women. It is also the second leading cause of death among women around the world, becoming the most fatal form of cancer. However, to detect and classify masses is a hard task even for experts. Therefore, due to medical experience, different diagnoses to an image are commonly found. The use of a computer assisted diagnosis is important to avoid misdiagnoses. In this work, we propose Fuzzy Morphological Extreme Learning Machines, with hidden layer kernel based on nonlinear morphological operators of erosion and dilation. The proposed methods were evaluated using 2.796 images from IRMA database, considering fat, fibroid, dense and extremely dense tissues. Zernike Moments and Haralick texture features are used as image descriptors. The proposed model classifies masses as benign, malignant or normal. Results shows comparison between Extreme Learning Machines using Sigmoid and Fuzzy Morphological Kernels, evaluated using classification rate and Kappa index. When using fuzzy morphological kernels, classification rate and Kappa value increases for most of cases analyzed.
According to theWorld Health Organization, breast cancer is the most common form of cancer in women. It is the second leading cause of death among women around the world, becoming the most fatal form of cancer. However, to detect and classify masses is a hard task even for experts. Consequently, due to medical experience, different diagnoses to an image are commonly found. Therefore, the use of a computer assisted diagnosis is important to avoid misdiagnoses. In this work, we propose Morphological Extreme Learning Machines, with hidden layer kernel based on nonlinear morphological operators of erosion and dilation. The proposed approach is evaluated using 2.796 images from IRMA database, considering fat, fibroid, dense and extremely dense tissues. Zernike Moments and Haralick texture features are used as image descriptors and the proposed model classifies the masses in benign, malignant or normal. Results shows comparison between Extreme Learning Machines using Sigmoid and Morphological Kernel, which are evaluated through classification rate and Kappa index. When using morphological kernels, the classification rate and Kappa value increases for most of cases analyzed.
Java vulnerabilities correspond to 91% of all exploits observed on the worldwide web. The present work aims to create antivirus software with machine learning and artificial intelligence and master in Java malware detection. Within the proposed methodology, the suspected JAR sample is executed to intentionally infect the Windows OS monitored in a controlled environment. In all, our antivirus monitors and considers, statistically, 6824 actions that the suspected JAR file can perform when executed. Our antivirus achieved an average performance of 91.58% in the distinction between benign and malware JAR files. Different initial conditions, learning functions and architectures of our antivirus are investigated. The limitations of commercial antiviruses can be supplied by intelligent antiviruses. Instead of blacklist-based models, our antivirus allows JAR malware detection preventively and not reactively as Oracle’s Java and traditional antivirus modus operandi.