Open AccessCBIR System for Lung Nodule Retrieval in CT Scans
Author(s) -
Ashwini Dasare,
Asst Prof
Publication year - 2019
Publication title -
international journal of innovative technology and exploring engineering
Language(s) - English
Resource type - Journals
ISSN - 2278-3075
DOI - 10.35940/ijitee.b1063.1292s19
Subject(s) - nodule (geology) , computer science , lung cancer , image retrieval , artificial intelligence , domain (mathematical analysis) , precision and recall , content based image retrieval , similarity (geometry) , pattern recognition (psychology) , radiology , medicine , image (mathematics) , mathematics , pathology , biology , paleontology , mathematical analysis
Lung cancer remains one of the fatal diseases with very high mortality rate in both men and women. Computer aided diagnostic systems have been contributing towards the enhancement of survival rate to a maximum extent. Most of such systems yield binary results, i.e. they classify whether a nodule is benign or malignant and they are computationally expensive. This paper proposes a methodology to build a Content Based Image Retrieval (CBIR) system that provides additional provision to the domain experts. Since the CBIR systems retrieve most similar images, this visual dimension will assist the budding and experience radiologist to assess the nodule information to greater detail. Nine visual and shape features are extracted for each nodule image collected from LIDC database and Minkowski distance measure is used for computing similarity. Experiments are conducted on 750 nodules out of which 375 are benign and 375 are malignant as identified by domain experts. Precision, recall and F measure metrics are considered to evaluate the methodology with achieved average values of 0.92, 0.82 and 0.86 respectively.