Image prototype similarity matching for lymph node hemopathology

Abstract

This paper describes general aspects of an automated expert system for Lymph Node Hemopathology, which utilizes methods of segmentation and classification for performing image prototype similarity matching (IPSM). The expert system consists of a set of representative prototype images of a large number of histologic features required to differentiate different lymph node pathologies. A queiy case, which may consist of one or more images, is compared against each prototype set and is assigned a degree of similarity by calculating a distance metric in a multidimensional feature space. Introductory motivation to this problem is presented together with technical details of the low-level segmentation algorithms utilized. As a representative application, results are presented for cases which are dominated by cytologic characteristics.

Lymph Node Image Segmentation Based on Improved FCM Clustering and Multi-Threshold

“…There are a lot of image segmentation algorithms, such as the classic edge extraction operator: Roberts, Sobel, Prewitt, LOG, Canny, etc. These operators are simpler, faster and more sensitive to noise interference [2] [3].The lymph node segmentation method based on the color attributes and fuzzy C-means was proposed by David N. Olivieri in 2000 [4]. Zhang Junhua and Wang Yuanyuan introduced a segmentation algorithm based on edge flow of improved gradient vector flow (GVF) to segment lymph node ultrasound images [5].…”