Region-of-interest retrieval in lung CT images using feature point maps

Abstract

In this paper, we present a novel method for retrieval of similar pathological cases in response to a query case in lung CT image database. A database of CT scan image slices of lung Adenocarcinoma is considered for experiments. The basic approach is to devise a framework, which lets a user input a 2D lung CT slice as a query image and output visually similar pathological images from the lung image case database. Feature point maps are used to represent the region-of-interest. Chamfer distance transform has been used as a measure to evaluate similarity for retrieval framework. The retrieval of similar pathological case slices in a quick time can aid the diagnosis and treatment of a new subject based on patient histories of the retrieved scans.