Radiogenomics-based cancer prognosis in colorectal cancer

Abstract

Radiogenomics aims at investigating the relationship between imaging radiomic features and gene expression alterations. This study addressed the potential prognostic complementary value of contrast enhanced computed tomography (CE-CT) radiomic features and gene expression data in primary colorectal cancers (CRC). Sixty-four patients underwent CT scans and radiomic features were extracted from the delineated tumor volume. Gene expression analysis of a small set of genes, previously identified as relevant for CRC, was conducted on surgical samples from the same tumors. The relationships between radiomic and gene expression data was assessed using the Kruskal-Wallis test. Multiple testing was not performed, as this was a pilot study. Cox regression was used to identify variables related to overall survival (OS) and progression free survival (PFS). ABCC2 gene expression was correlated with N (p = 0.016) and M stages (p = 0.022). Expression changes of ABCC2, CD166, CDKNV1 and INHBB genes exhibited significant correlations with some radiomic features. OS was associated with Ratio 3D Surface/volume (p = 0.022) and ALDH1A1 expression (p = 0.042), whereas clinical stage (p = 0.004), ABCC2 expression (p = 0.035), and Entropy_{GLCM_E} (p = 0.0031), were prognostic factors for PFS. Combining CE-CT radiomics with gene expression analysis and histopathological examination of primary CRC could provide higher prognostic stratification power, leading to improved patient management.

Figure 1

Kaplan-Meier analysis of PFS with Cox models combining Stage III, ABCC2 expression and Entropy_{GLMC E} led to increased stratification power.

Figure 2

Workflow of prognosis model construction. A colorectal tumor is delineated in every slice and validated by an experienced physician. This allows creation of a 3D representation of the tumor. Radiomic features (intensity, shape, texture) are extracted from this delineated tumor, and integrated with clinical, histopathology and genomic/gene expression data.