Associations between MRI radiomics analysis and tumor-micro milieu in uterine cervical cancer.
The complex interactions of the tumor micromilieu could be reflected by magnetic resonance imaging (MRI) when analyzed with the radiomics approach. For several tumor entities, it has been shown that radiomics derived from MRI can reflect important characteristics of the tumors. The present study investigated the association radiomics derived from MRI images and histopathological features in uterine cervical cancer.
The MRI before any treatment was used to extract the radiomics features of T1- and T2-weighted images. The biopsy specimens were stained for Ki 67, e-cadherin, vimentin, programmed-death ligand 1, and tumor-infiltrating lymphocytes (TIL, all CD45 positive cells). Tumor-stroma ratio (TSR) was calculated on routine H&E specimen. Spearman's correlation analysis and discrimination analyses were performed as statistical analyses.
The patient sample was comprised of 89 female patients with a mean age of 49.3 years ± 14.6 (range 27-77 years) with squamous cell cervical carcinoma. "Kurtosis" derived from T1-weighted images after contrast media application correlated with the Ki-67 index (r = 0.28, p = 0.02). "WavEnHL_s-4" derived from T2-weighted images and "S(1.0)Contrast" derived from T1-weighted images after contrast media application showed correlations with TSR (r = - 0.24, p = 0.04, each). Several associations were identified between the radiomics features with immune scores defined by programmed-death ligand 1, the highest correlation showed Teta1 derived from T2-weighted images with the combined positive score (r = - 0.38, p < 0.01). There were several associations with vimentin expression, the highest showed "Variance" derived from T1-weighted images after contrast media application (r = 0.46, p < 0.01).
Radiomics features derived from MRI can reflect tumor characteristics of UCC. Especially immune-related features were reflected by the MRI texture features. Proliferation potential, composition of the extracellular matrix and tumor-stroma ratio were also significantly associated with radiomics features. These presented results need to be evaluated in an independent cohort to test their stability.
The MRI before any treatment was used to extract the radiomics features of T1- and T2-weighted images. The biopsy specimens were stained for Ki 67, e-cadherin, vimentin, programmed-death ligand 1, and tumor-infiltrating lymphocytes (TIL, all CD45 positive cells). Tumor-stroma ratio (TSR) was calculated on routine H&E specimen. Spearman's correlation analysis and discrimination analyses were performed as statistical analyses.
The patient sample was comprised of 89 female patients with a mean age of 49.3 years ± 14.6 (range 27-77 years) with squamous cell cervical carcinoma. "Kurtosis" derived from T1-weighted images after contrast media application correlated with the Ki-67 index (r = 0.28, p = 0.02). "WavEnHL_s-4" derived from T2-weighted images and "S(1.0)Contrast" derived from T1-weighted images after contrast media application showed correlations with TSR (r = - 0.24, p = 0.04, each). Several associations were identified between the radiomics features with immune scores defined by programmed-death ligand 1, the highest correlation showed Teta1 derived from T2-weighted images with the combined positive score (r = - 0.38, p < 0.01). There were several associations with vimentin expression, the highest showed "Variance" derived from T1-weighted images after contrast media application (r = 0.46, p < 0.01).
Radiomics features derived from MRI can reflect tumor characteristics of UCC. Especially immune-related features were reflected by the MRI texture features. Proliferation potential, composition of the extracellular matrix and tumor-stroma ratio were also significantly associated with radiomics features. These presented results need to be evaluated in an independent cohort to test their stability.
Authors
Meyer Meyer, Leonhardi Leonhardi, Höhn Höhn, Kabbani Kabbani, Zimmermann Zimmermann, Borggrefe Borggrefe, Surov Surov
View on Pubmed