Published online Jan 27, 2022. doi: 10.4254/wjh.v14.i1.244
Peer-review started: April 25, 2021
First decision: June 13, 2021
Revised: June 15, 2021
Accepted: December 2, 2021
Article in press: December 2, 2021
Published online: January 27, 2022
Radiomics is a rapidly growing field of radiological research. In radiomics, computed tomography (CT) texture analysis quantifies tissue heterogeneity and has shown promise in predicting pathological features, the overall survival and the response to therapy in oncology. In the last few years a few studies have reported that texture analysis can be helpful in predicting the response to chemotherapy for colorectal liver metastases, but the results have been heterogeneous.
In previously published texture analysis studies on the first-line chemotherapy response of colorectal liver metastases (CRLMs), necrotic CRLMs were not clearly excluded. Thicker CT slice reconstructions were utilized in most studies which could have influenced the radiomics results due to partial voxel artefacts. Limited first and second order texture features were also analyzed in previous studies.
The aim of this study was to identify new predictive imaging biomarkers in patients with non-necrotic CRLMs who received first-line cytotoxic chemotherapy. CT texture analysis was performed on non-necrotic CRLMs utilizing 1.25 mm portal venous phase CT reconstructions. We also assessed a larger range of first and second order texture features.
A total of 236 patients with CRLMs who received first-line cytotoxic chemotherapy in our private institution from March 2012 to May 2020 were retrospectively identified on our radiology information system. There were various inclusion and exclusion criteria and the final study cohort consisted of 29 patients. Multiple first and second order texture features were analyzed with the SOPHiA Radiomics software to identify predictive biomarkers in the responding CRLMs.
Our study identified a few new texture features and a promising radiomics signature which are significantly associated with the response of CRLMs to first-line cytotoxic chemotherapy. In univariable analysis eight texture features of the responding non-necrotic CRLMs were associated with treatment response, but due to strong pairwise correlations among some of the features, only two features namely minimum histogram gradient intensity and long run low grey level emphasis were included in the multiple analyses and final radiomics signature. The results of this study were unique but need to be validated and confirmed on larger patient cohorts.
Future radiomics studies should attempt to quantify the difference in the texture analysis results of necrotic vs non-necrotic CRLMs utilizing different CT slice reconstructions in the same study cohort to compare the predictive value of texture analysis. These factors may partially account for the heterogeneous results which have been reported in the last few years. To allow for the better comparison between radiomics studies we should work towards the standardization of study designs, interscanner differences, acquisition parameters, analysis algorithms, the feature extraction techniques, analysis methodologies and the group of texture features which should be evaluated based on the different types of cancer.
The preliminary results of our study need to be validated and confirmed on larger patient cohorts. Further investigations are required to determine if the predictive texture features have any prognostic value and are linked to the KRAS mutation status of CRLMs. Standardization of radiomics studies is required to compare the texture analysis results of different studies.