RSNA 2014 

Abstract Archives of the RSNA, 2014


CT Texture Analysis of Histologically Proven Benign and Malignant Lung Lesions

Scientific Papers

Presented on December 2, 2014
Presented as part of SSG07: Informatics (3D, Quantitative and Advanced Visualization)

 Trainee Research Prize - Resident


Roberto Lo Gullo MD, Presenter: Nothing to Disclose
Mannudeep K. S. Kalra MD, Abstract Co-Author: Nothing to Disclose
Atul Padole MD, Abstract Co-Author: Nothing to Disclose
Alexi Otrakji MD, Abstract Co-Author: Nothing to Disclose
Jo-Anne O. Shepard MD, Abstract Co-Author: Consultant, Agfa-Gevaert Group
Subba Rao Digumarthy MD, Abstract Co-Author: Nothing to Disclose
Sarabjeet Singh MD, Abstract Co-Author: Research Grant, Siemens AG Research Grant, Toshiba Corporation Research Grant, General Electric Company Research Grant, Koninklijke Philips NV
Ranish Deedar Ali Khawaja MD, Abstract Co-Author: Nothing to Disclose


To assess differences in CT texture of histologically proven benign and primary malignant lung lesions.


Our ongoing IRB approved study included 62 patients (28M; 34F; mean age 64 years; range27-88) who underwent CT guided biopsy and had definitive diagnosis on pathology [22 benign (9 tumor and tumor-like and 13 acute infections), 40 malignant (17 well differentiated and 23 poorly differentiated primary lung adenocarcinomas)], over a period of three years. All mediastinal, chest wall, pleural lesions and pulmonary metastases were excluded. CT texture analysis of all biopsied lesions was performed on the pre-biopsy non-contrast CT images using a commercially available software (TexRAD limited, UK). For each patient, regions of interest were drawn on a single image with largest lesion dimensions. Areas of cavitation containing air were excluded. The features analyzed included mean HU values, percent positive pixels (PPP), mean value of positive pixels (MPP), standard deviation (SD), normalized SD, skewness, kurtosis, and entropy. Data were analyzed using two tailed unpaired non parametric T-test with Welch correction


There were significant differences between CT texture of well- and poorly-differentiated primary lung cancers (p<0.0001) with respect to SD (respectively 33.8±5.9 vs 29.2 ±6.5), normalized SD (6.5±2.5 vs 4.2±1.2), Kurtosis (-0.5±0.4 vs 0.1±0.6) and PPP (0.7±0.1vs0.8±0.1). There is significant difference in normalized SD among the benign lesions between infection and tumor/tumor like lesions (4.3±1.4 vs 7±2.1; p< 0.01). The CT texture of benign tumors and tumor like lesions was similar to well differentiated lung cancer.


CT texture analysis can separate well diffentiated and poorly differentiated primary lung adenocarcinomas. There is a significant difference in SD between tumor/tumor-like lesions and infections among the benign lesions.


CT texture analysis can help characterize malignant and benign lung lesions and can separate well differentiated from poorly differentiated cancers, which may help in selecting treatment options.

Cite This Abstract

Lo Gullo, R, Kalra, M, Padole, A, Otrakji, A, Shepard, J, Digumarthy, S, Singh, S, Khawaja, R, CT Texture Analysis of Histologically Proven Benign and Malignant Lung Lesions.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL.