Radiomics analysis in gastrointestinal imaging: a narrative review

Background and Objective: To present an overview of radiomics radiological applications in major gastrointestinal oncological non-oncologic diseases, such as colorectal cancer, pancreatic cancer, gastro- oesophageal cancer, gastrointestinal stromal tumor (GIST), hepatocellular carcinoma (HCC), intrahepatic cholangiocarcinoma (ICC), and non-oncologic diseases, such as liver fibrosis, nonalcoholic steatohepatitis, and inflammatory bowel disease. Methods: A search of PubMed databases was performed for the terms “radiomic”, “radiomics”, “liver”, “small bowel”, “colon”, “GI tract”, and “gastrointestinal imaging” for English articles published between January 2013 and July 2022. A narrative review was undertaken to summarize literature pertaining to application of radiomics in major oncological and non-oncological gastrointestinal diseases. The strengths and limitation of radiomics, as well as advantages and major limitations and providing considerations for future development of radiomics were discussed. Key Content and Findings: Radiomics consists in extracting and analyzing a vast amount of quantitative features from medical datasets, Radiomics refers to the extraction and analysis of large amounts of quantitative features from medical images. The extraction of these data, integrated with clinical data, allows the construction of descriptive and predictive models that can build disease-specific radiomic signatures. Texture analysis has emerged as one of the most important biomarkers able to assess tumor heterogeneity and can provide microscopic image information that cannot be identified with the naked eye by radiologists. Conclusions: Radiomics and texture analysis are currently under active investigation in several institutions worldwide, this approach is being tested in a multitude of anatomical areas and diseases, with the final aim to exploit personalized medicine in diagnosis, treatment planning, and prediction of outcomes. Despite promising initial results, the implementation of radiomics is still hampered by some limitations related to the lack of standardization and validation of image acquisition protocols, feature segmentation, data extraction, processing, and analysis