Automatic Segmentation of Chondroblastoma Using x ray Images

Abstract

Bone disease prediction using the medical images is the most necessary task in the real world environment, which needs to process the images well to attain an improved prediction. The analysis of bone disease-related data sets using various computer-aided methods helps to obtain the enhanced prediction of diseases. The death rate due to the tumour has been increasing enormously over the past three decades. Chondroblastoma is one such bone tumour which is locally aggressive and typically affects the epiphysis of long bones. It is very aggressive and life-threatening, which must be promptly diagnosed and cured to prevent mortality. newlineUsually, the bone tumour analysis and manual segmentation of tumour part is a tedious and time-consuming task. Due to the presence of noise and other imaging drawbacks, a completely automated, computer-aided system has become a challenging task in medical image processing. There exist different modalities of medical imaging systems like X-rays, Ultrasounds, Magnetic Resonance Imaging (MRI), Computed Tomography (CT) and Positron Emission Tomography (PET). Most of these modalities usually suffer from noise and other sampling issues. Moreover, all these modalities depend mainly on the interpretation of the radiologist s knowledge, experience, and are mostly time-dependent. This work proposes an approach to detect Chondroblastoma from X-ray images using medical image processing techniques. X-ray imaging is one of the most popular and commonly used methods to identify and locate the tumour. newlineThis thesis proposes a fully automatic method for Chondroblastoma segmentation from X-ray images by using a modified region-based method. The recommended system consists of two stages, such as pre-processing and segmentation. In pre-processing, the noise and other capture issues present in the image are removed. Segmentation of Chondroblastoma is developed by clustering, edge-based and region-based approaches. In clustering-based segmentation, K-Means and Fuzzy C Means (FCM) algorithms are used.