Skip to main content

Table 2 Studies investigating the use of CT texture analysis in diagnosis, treatment response assessment, and as a prognostic tool

From: Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice?

Diagnosis and characterisation Method Study findings Author, year
Diagnosis
Lung
Pulmonary nodules Fractal analysis 3D fractal dimension was higher in organizing pneumonias/tuberculomas than carcinomas/hamartomas (p < 0.001) and higher in adenocarcinomas than squamous cell (p < 0.05) Kido et al., 2002 [37]
Bronchoalveolar carcinoma vs. non-bronchoalveolar carcinoma Fractal analysis Fractal dimension higher for bronchoalveolar carcinomas (2.38 ± 0.05/2.16 ± 0.01) than non- bronchoalveolar carcinomas (2.19 ± 0.05/2.06 ± 0.01 internal/peripheral; p < 0.0001) Kido et al., 2003 [36]
Lung cancer Fractal analysis Fractal dimension was higher for stage III and IV cancers than stage I (2.046 vs. 1.534). 83.8 % of stage IV tumors were classified as aggressive with a threshold of 1.913 Al-Kadi et al., 2008 [5]
Liver
Hepatic tumors Texture analysis Autocovariance function differed between malignant (HCC and colorectal metastases) and benign lesions. Sensitivity of 75.0 % and specificity of 88.1 % were achieved with the proposed diagnostic system Huang et al., 2006 [38]
GI tract
Colorectal cancer Fractal analysis Fractal dimension and abundance were higher in colon cancer than normal bowel: mean (SD) 1.71(0.07) vs. 1.61(0.07) for dimension and 7.82(0.62) vs. 6.89 (0.47) for abundance (P ≤ 0.001) Goh et al., 2007 [8]
Colorectal cancer Texture analysis Fractal dimension is higher for metastatic nodes Cui et al., 2011 [39]
Brain
Glioma Texture analysis Coarse texture entropy >5.2 had a sensitivity and specificity of 76 % and 82 %, respectively; uniformity <0.025 had a sensitivity and specificity of 64 % and 95 %, respectively, for high-grade tumors Skogen et al., 2011 [40]
Response assessment
Metastatic renal cell carcinoma Texture analysis Percentage change in coarse texture uniformity of ≤ −2 % after 2 cycles of TKI correlated with shorter time to progression Goh et al., 2011 [41]
Prognosis assessment
Liver texture in patients with colorectal cancer but no known metastases Texture analysis Coarse texture entropy correlated with hepatic perfusion index
(r = −0.503978, p = 0.007355) and survival (r = 0.489642, p = 0.009533). Hypothesized texture features may reflect vascular changes associated with micrometastases. Entropy <2.0 identified patients who died with 100 % sensitivity, 65 % specificity
Ganeshan et al., 2007 [42]
Colorectal cancer metastases Texture analysis Uniformity at texture ratios of 1.5/2.5 and 2.0/2.5 were significant OS prognostic factors (p < 0.005) Miles et al., 2009 [43]
Liver texture in patients with colorectal cancer Texture analysis Fine texture entropy of ≤0.0807 between 26–30 s after contrast injection highlighted node-positive patients with 100 % sensitivity, 71 % specificity. HPI did not vary significantly between node-negative and -positive patients Ganeshan et al., 2011 [32]
Esophageal cancer Texture analysis Unenhanced CT component of PET-CT Ganeshan et al., 2012 [33]
  Greater heterogeneity in higher stage tumors. Coarse uniformity was a significant OS prognostic factor (OR = 4.56, 95 % CI 1.08–18.37, p = 0.039)  
NSCLC Texture analysis Coarse texture uniformity <0.624 was a poor prognostic factor Ganeshan et al., 2011 [34]