Journal
CANCERS
Volume 14, Issue 19, Pages -Publisher
MDPI
DOI: 10.3390/cancers14194871
Keywords
radiomics; reproducibility; oncologic imaging; evidence-based medicine; research quality
Categories
Ask authors/readers for more resources
Radiomics has the potential to enhance the value of medical images for oncologic patients, but its clinical application faces challenges and requires a rigorous methodology. This review provides guidelines to facilitate state-of-the-art radiomics research.
Simple Summary Radiomics could increase the value of medical images for oncologic patients, allowing for the identification of novel imaging biomarkers and building prediction models. Unfortunately, despite the many promises and encouraging findings, the translation of radiomics into clinical practice appears as a distant goal. Indeed, challenges such as generalizability and reproducibility are slowing down the process and must be faced with a rigorous and robust radiomic methodology. In this review, we turn the spotlight to the methodological complexity of radiomics, providing an outline of dos and don'ts aimed at facilitating state-of-the-art research. Imaging plays a crucial role in the management of oncologic patients, from the initial diagnosis to staging and treatment response monitoring. Recently, it has been suggested that its importance could be further increased by accessing a new layer of previously hidden quantitative data at the pixel level. Using a multi-step process, radiomics extracts potential biomarkers from medical images that could power decision support tools. Despite the growing interest and rising number of research articles being published, radiomics is still far from fulfilling its promise of guiding oncologic imaging toward personalized medicine. This is, at least partly, due to the heterogeneous methodological quality in radiomic research, caused by the complexity of the analysis pipelines. In this review, we aim to disentangle this complexity with a stepwise approach. Specifically, we focus on challenges to face during image preprocessing and segmentation, how to handle imbalanced classes and avoid information leaks, as well as strategies for the proper validation of findings.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available