Aims and objectives
■ Differentiating benign from malignant pulmonary nodules is critical in the management of patients with pulmonary nodule.
■Recent studies have shown that incorporating lung nodule characteristics identified on CT scans such as size,
texture,
growth rate and contrast enhancementcan improve the accuracy of predicting the risk of malignancy.1-4
■ The purpose of this study was to investigate the use of textural features and machine learning methods in comparison to CT contrast enhancement and metabolic activity in differentiating benign from malignant pulmonary nodules.
Methods and materials
■ 85 pulmonary nodules examined with dynamic contrast-enhanced chest CT and PET/CT scans were included.
■ 58 were diagnosed by histology and 27 by radiological follow-up.
38 (44.7%) were malignant.
■ Contours of the pulmonary nodules were drawn manually on the one-minute post-contrast scan and propagated using deformable image registration (Mirada XD,
Mirada Medical,
Oxford,
UK).
■The SUVmax,
enhancement,
textural features,
sub-volumes inside and outside the contours were computed.
■Separate Gaussian models learnt for malignant and benign populations on subsets of 2-3 features were used...
Results
■ 70/85 (82%) nodules were correctly classified as malignant/benign using the three-feature Gaussian texture model in a leave-one-out paradigm.
■ In comparison,
51/85 and 43/85 (60% and 51%) were correctly classified as malignant/benign using enhancement of greater than 20 HU and maximum standardized uptake value (SUVmax) of 2.5,
respectively.
■ The 3 optimal features for the texture model were “fractalness” of the nodule shape,
the 33rd percentile for brightness inside the nodule,
and mean brightness outside the nodule,
measured 3 minutes post-contrast.
■ ROC-based analysis...
Conclusion
Textural and image analysis using machine learning methods may help differentiate benign from malignant pulmonary nodules along,
with metabolic activity and enhancement.
Personal information
J.
S.
Z.
Lee,1 L.
C.
Pickup,2 A.
Larrue,2M.
Gooding,2T.
Kadir,2F.
V.
Gleeson1
1Oxford University Hospitals NHS Trust,
UK
2Mirada Medical,
UK
References
Ganeshan B,
Panayiotou E,
Burnand K,
Dizdarevic S,
Miles K.
Tumour heterogeneity in non-small cell lung carcinoma assessed by CT texture analysis: a potential marker of survival.
Eur Radiol.
2012 May;22(4):796–802.
MacMahon H,
Austin JHM,
Gamsu G,
Herold CJ,
Jett JR,
Naidich DP,
et al.
Guidelines for management of small pulmonary nodules detected on CT scans: a statement from the Fleischner Society.
Radiology.
2005 Nov;237(2):395–400.
Mikita K,
Saito H,
Sakuma Y,
Kondo T,
Honda T,
Murakami S,
et al.
Growth rate of lung cancer recognized...