Open AccessFeature based analyses of lung nodules from computed tomography (CT) images
Author(s) -
Md. Anwar Hussain,
Lakshipriya Gogoi
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1020/1/012007
Subject(s) - computer science , feature (linguistics) , artificial intelligence , computed tomography , matlab , nodule (geology) , segmentation , lung cancer , lung cancer screening , software , field (mathematics) , tomography , image processing , radiology , pattern recognition (psychology) , image (mathematics) , medicine , pathology , mathematics , paleontology , philosophy , linguistics , biology , pure mathematics , programming language , operating system
Among various lung image modalities, CT (computed tomography) images have been found most suitable and widely used for detection of small lung nodules. Although an expert radiologist can analyse these images quite perfectly, however an efficient CADe (computer-aided detection ) system capable of detecting pulmonary nodules automatically may be of great help considering the large number of CT images, a radiologist needs to analyse a day in recent years. A few CADe systems have already been tested within lung cancer screening trial which have enjoyed mixed results. So, the field has enough voids to be filled and research towards development of novel and efficient CADe systems has become an interesting, but challenging field of research. We too have been trying to develop CADe systems that can analyse CT images to detect and identify various lung nodules. In the process, we are reporting herein one of our development for the same. Our CADe system has maintained its 81.25% accuracy in detecting malignant nodules. Hand-crafted features of the selected lung CT images were used in the study. The study is emphasised on developing efficient CADe systems capable of detecting solid nodules (size > 2 mm) at different locations, whether isolated, juxtapleural or juxtavascular nodules. While MATLAB was used to carry out pre-processing, segmentation and nodule detection, testing of datasets was done by a MLP (multilayer perceptron) network by using WEKA (Waikato Environment for Knowledge Analysis) software.