A Novel Approach For Diagnosis And Prognostic Analysis Of Abdominal Aortic Aneurysm With Digital Image Processing

Abstract

Abdominal aorta aneurysm is preternatural focal dilation of the aorta. It occurs somewhere between the aortic hiatus diaphragm from the segmentation of common iliac blood vessels. AAAs typically grows at a rate of around 2.6-3.2 mm/ year. If it tends to exceed or grow faster, it results in a risk of rupture so early prediction of AAA helps to diagnose in an efficient way. To detect it in an effective way, image processing technique is applied to have a highly qualified result in diagnosis. Image processing plays a prominent role in estimating images with respect to various criteria. While processing the image, image segmentation is a crucial step. It performs a main role throughout the process. To provide an accurate result, medical image processing technique is not much efficient to determine the size and position of affected region of aneurysm. So, computer aided diagnostic tools were preferred. A clear scrutinization and learning are required by a specialist to recognize the problem in aneurysm. The obstructions, which was involved in aneurysm has been handled through proposed automated algorithms. It undergoes a process of several phases, initiating from the process of filtering and extends to the feature extraction technique and finally classification was performed using effective classifier. Initially preprocessing is performed to get a high-quality image. Adaptive median filter minimizes the problem faced by notch and median filter. The problem arises because of the dependency and it requires large number of parameters to perform computation. Adaptive median filter works well in filtering out the noise effectively and it also has a capability in achieving very fine details in the image and the next phase is segmentation. In segmentation, the image gets v grouped into samples based upon the pixel. Analogizing with exudate, watershed transform and HLSFMM segmentation, artificial neural network-based segmentation performs much effective and helps to extract the features in an accurate manner. For feature extraction and optimization, particle swarm optimization is compared with GRCM and fuzzy C-means clustering. Even though, fuzzy is good in handling uncertainties, it completely dependent upon human expertise and it consumes more time in performing computation but particle swarm optimization performs computation better and it selects the best feature effectively and final phase is to perform classification between trained and test set data. Grey neural network classifier works better in analogizing with SVM and DNN classifier. SVM classifier is prone to the problem of over fitting which results in poor performance and DNN requires the parallel processing power. The proposed method was tested for 200 samples and it generates the promising result for the specialist to diagnose the problem and treat the patients accordingly. newline