Detection of lung nodules using unsupervised machine learning method

  • Raj Kishore
  • , Manoranjan Satpathy
  • , D. K. Parida
  • , Zohar Nussinov
  • , Kisor K. Sahu

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Machine learning methods are now becoming a popular choice in many computer-aided bio-medical image analysis systems. It reduces the efforts of a medical expert and helps in making correct decisions. One of the main applications of such systems is the early detection of lung cancerous nodules using Computed Tomography (CT) scan images. Here, we have used a new method for automated detection of lung cancerous/non-cancerous nodules. It is a modularity maximization based graph clustering method. The clustering is done based on the different region’s grayscale values of the CT scan images. The clustering algorithm is capable of detecting nodules of size as small as 4 pixels in two dimension (2D) or 9 voxels in three dimensional (3D) data. The advantage of nodule detection is that it can be used as an extra feature for many supervised learning algorithms especially for those Convolutional Neural Networks (CNN) based architectures where pixel-wise segmentation of data might be required.

Original languageEnglish
Title of host publicationComputational Vision and Bio-Inspired Computing, ICCVBIC 2019
EditorsS. Smys, João Manuel R.S. Tavares, Valentina Emilia Balas, Abdullah M. Iliyasu
PublisherSpringer
Pages463-471
Number of pages9
ISBN (Print)9783030372170
DOIs
StatePublished - 2020
Event3rd International Conference on Computational Vision and Bio Inspired Computing, ICCVBIC 2019 - Coimbatore, India
Duration: Sep 25 2019Sep 26 2019

Publication series

NameAdvances in Intelligent Systems and Computing
Volume1108 AISC
ISSN (Print)2194-5357
ISSN (Electronic)2194-5365

Conference

Conference3rd International Conference on Computational Vision and Bio Inspired Computing, ICCVBIC 2019
Country/TerritoryIndia
CityCoimbatore
Period09/25/1909/26/19

Keywords

  • CNN
  • CT scan
  • Graph clustering
  • Lung cancer
  • Machine learning
  • Modularity
  • Segmentation

Fingerprint

Dive into the research topics of 'Detection of lung nodules using unsupervised machine learning method'. Together they form a unique fingerprint.

Cite this