Abstract

Many a times, the quality of Magnetic Resonance Imaging (MRI) based images get affected by the machines that are utilized to compute them. Hence, in several cases, it is necessary to process each image in order to obtain a corresponding better image so that the patient can be treated with ease and accuracy. According to traditional methods, this objective can be achieved by contrast enhancement or histogram stretching, etc. This paper give a naive approach that can enhance the MRI image, classify the brain tumor type and segment the tumor region for better analysis. For the enhancement, the concept of Discrete Wavelet is used in which an image is decomposed in four sub-bands and each sub-band is processed separately after which Inverse Discrete Wavelet Transform (IDWT) is applied in order to acquire a better quality image. Convolutional Neural Network (CNN) is used to classify the type of tumor present in MRI image. Lastly, after the tumor is detected, it is segmented by using a very efficient method known as K-means Clustering.

中文翻译：

---

用于脑肿瘤分类的磁共振图像增强和分割模型

摘要

很多时候，基于磁共振成像（MRI）的图像的质量会受到用于计算它们的机器的影响。因此，在几种情况下，必须处理每个图像以获得相应的更好的图像，以便可以容易且准确地治疗患者。根据传统方法，可以通过增强对比度或直方图拉伸等方法来实现此目标。本文提供了一种可以增强MRI图像，对脑肿瘤类型进行分类并对肿瘤区域进行分割以进行更好分析的幼稚方法。为了增强效果，使用了离散小波的概念，其中将图像分解为四个子带，并分别处理每个子带，然后再应用离散小波逆变换（IDWT）以获得更好的图像质量。卷积神经网络（CNN）用于对MRI图像中存在的肿瘤类型进行分类。最后，在检测到肿瘤后，使用称为K-均值聚类的非常有效的方法对其进行分割。