A Brain Tumor Segmentation and Detection Technique Based on Birch and Marker Watershed

Today, image segmentation plays a vital role in many crucial areas. This paper will cover the important role that image segmentation plays in medical imaging issues. Image segmentation is the process of dividing a random image into groups or partitions; the main goal here is to separate the objects from the background, to make it easier to analyze each of them individually, like in our case when the purpose is to detect and segment the tumor. In the upcoming sections, we’ll be presenting a new medical imaging technique for detecting brain tumors in MRI images. For this, we will use two algorithms. The first one is Birch which is a fast hierarchical method of grouping and reducing data. The second algorithm is the segmentation by the Watershed, these two algorithms will be explained further in detail. The first step is to segment the input MRI image using the Birch method, then the resulting image will be launched into the second algorithm which is the Watershed. The proposed method has been applied and tested on 155 images from the Kaggle dataset to detect abnormalities such as tumors. According to the results, we can say that it gives a satisfactory and efficient detection result. To evaluate the performance of our proposed method, we used a very known set of parameters such as the dice coefficient that reached 96%, the sensitivity achieved 99%, in specificity we obtained 98, and a Jaccard similarity of 94%. All the given evaluating metrics achieved satisfactory results and good segmentation of the brain tumor. Furthermore, our proposed approach was compared to the SOTA methods, and the obtained results outperformed the state-of-the-art brain tumor extraction by 5%.