Fuzzy Image Processing for Diagnosing In ammation in Pulmonary Biopsies

Abstract

This paper proposes a new approach to diagnose the degree of in ammation in digital images of pulmonary biopsies, provided by a digital camera through a microscope. Diagnosing is done by detecting thick epithelium cell layers around the vessels and bronchus in tissue images. For analyzing the complex images of tissue, a fuzzy image processing procedure consisting of ve main stages is presented. The rst stage is decreasing the complexity of the images by using image pre-processing methods for enhancement and smoothing the image with a Gaussian low pass lter in order to highlight important details and ignore the unnecessary parts of the image. The second stage is segmentation by using a fuzzy c-means clustering algorithm and fuzzy canny edge detection. This step works as a data reduction method as well as object recognition. Feature extraction, the third stage, will be done by using a fuzzy Hough transform. After extracting features such as bronchioles and vessels from the image, the fourth stage will be analysis and reasoning by a fuzzy inference system, which is a hybrid of the Mamdani and Logical modeling system with a Yager parametric operator. The last stage is tuning system parameters and the learning process with a feed forward neural network. The output of the proposed algorithm is the degree of in ammation inferred by the fuzzy inference system. The proposed approach is user friendly with low computational time and the results are more precise, reliable and acceptable to experts and physicians.