Associations between radiologist-defined semantic and automatically computed radiomic features in non-small cell lung cancer

Abstract

Tumor phenotypes captured in computed tomography (CT) images can be described qualitatively and quantitatively using radiologist-defined "semantic" and computer-derived "radiomic" features, respectively. While both types of features have shown to be promising predictors of prognosis, the association between these groups of features remains unclear. We investigated the associations between semantic and radiomic features in CT images of 258 non-small cell lung adenocarcinomas. The tumor imaging phenotypes were described using 9 qualitative semantic features that were scored by radiologists, and 57 quantitative radiomic features that were automatically calculated using mathematical algorithms. Of the 9 semantic features, 3 were rated on a binary scale (cavitation, air bronchogram, and calcification) and 6 were rated on a categorical scale (texture, border definition, contour, lobulation, spiculation, and concavity). 32-41 radiomic features were associated with the binary semantic features (AUC = 0.56-0.76). The relationship between all radiomic features and the categorical semantic features ranged from weak to moderate (|Spearmen's correlation| = 0.002-0.65). There are associations between semantic and radiomic features, however the associations were not strong despite being significant. Our results indicate that radiomic features may capture distinct tumor phenotypes that fail to be perceived by naked eye that semantic features do not describe and vice versa.

Figure 1

Association between the binary semantic and unfiltered radiomic features assessed with the area under the ROC curve (AUC). *Indicates a significant association (q-value ≤ 0.05). “Rand.” = random association (AUC = 0.50). “Prop.” and “Inv. Prop.” indicate direct and inverse proportionality, respectively.

Figure 2

Tumors with and without cavitation. (a) Tumor without cavitation (b) Tumor with minor Cavitation (c) Tumor with major Cavitation. The arrow indicates the location of the tumor.

Figure 3

Associations between the binary semantic and unfiltered radiomic features assessed with the area under the ROC curve (AUC). *Indicates a significant association (q-value ≤ 0.05). “Rand.” = random association (AUC = 0.50). “Prop.” And “Inv. Prop.” indicate direct and inverse proportionality, respectively. Wv = Wavelet. LoG = Laplacian of Gaussian.

Figure 4

Association between the six categorical semantic and ten unfiltered radiomic features assessed with Spearman coefficient correlation. *Indicates that the association was significant (q-value ≤ 0.05).

Figure 5

Associations between the categorical semantic and unfiltered radiomic features assessed with Spearman coefficient correlation. *Indicates a statistically significant association (q-value ≤ 0.05). Wv = Wavelet. LoG = Laplacian of Gaussian.

Figure 6

Tumors with different border definitions. (a) Tumor with a well-defined border (score = 1). (b) Tumor with neither a well- or poorly-defined border (score = 2). (c) Tumor with a poorly-defined border (score = 3). The arrow indicates the location of the tumor.