Computer-aided detection of lung nodules on multidetector row computed tomography using three-dimensional analysis of nodule candidates and their surroundings

Sumiaki Matsumoto, Yoshiharu Ohno, Hitoshi Yamagata, Daisuke Takenaka, Kazuro Sugimura

Research output: Contribution to journalArticlepeer-review

21 Citations (Scopus)

Abstract

Purpose. We have been developing a computer-aided detection (CAD) system for lung nodules on multidetector row computed tomography (MDCT). The scheme for nodule detection in this system is featured by three-dimensional analysis for nodule detection in nodules and their surroundings, which is designed to discriminate nodules from blood vessels. The purpose of this study was to evaluate the CAD system. Materials and methods. MDCT images from 30 patients with lung nodules were read twice, 3 weeks apart by a chest radiologist to detect noncalcified nodules of ≥4 mm. The first reading was without CAD, and the second reading was with CAD. Based on the reference standard later determined by another chest radiologist, the sensitivity of the former chest radiologist without or with CAD was obtained; the sensitivity and false-positive rate of the system alone were also obtained. Results. The reference standard consisted of 66 nodules. The sensitivity of the chest radiologist was 77% (51/66) without CAD and 91% (60/66) with CAD, showing a significant improvement. The CAD system alone showed a sensitivity of 79% (52/66) with the false-positive rate of 4.5 per patient. Conclusion. Although preliminary, these results indicate the efficacy of the CAD system.

Original languageEnglish
Pages (from-to)562-569
Number of pages8
JournalRadiation Medicine - Medical Imaging and Radiation Oncology
Volume26
Issue number9
DOIs
Publication statusPublished - 11-2008
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Radiation
  • Radiology Nuclear Medicine and imaging
  • Oncology

Fingerprint

Dive into the research topics of 'Computer-aided detection of lung nodules on multidetector row computed tomography using three-dimensional analysis of nodule candidates and their surroundings'. Together they form a unique fingerprint.

Cite this