Word of the Day: Radiogenomics
Significance of the Terminology of Today.
In recent years a new direction in cancer research has emerged that focuses on the relationship between imaging phenotypes and genomics. This direction is referred to as radiogenomics or imaging genomics.
Increasingly, even casual readers of the scientific literature are encountering such terms as radiogenomics, imaging genomics, and radiomics, which happen to be its three faces.
‘Radiogenomics’ is most frequently used to refer to the relationship between the imaging characteristics of a disease (a.k.a. imaging phenotype or radiophenotype) with its gene expression patterns, gene mutations, and other genome-related characteristics.
‘Imaging genomics’ refers to a research effort aimed at finding the relationship between imaging phenotypes and gene expression patterns which include expressions of individual genes as well as measures that summarize expressions of specific gene subsets.
‘Radiomics’ focuses on the methodology used in the analysis. Specifically, the radiomics paradigm proposes extraction of a large number of quantitative features from images using computer algorithms. The extracted features can then either be related to other data of interest, including patient outcomes, or to genomic characteristics.
Radiogenomics offers a practical way to leverage limited and incomplete data to generate knowledge that might lead to improved decision making in oncology. Its significance increases manifold due to practical limitations of currently available data that often lack complete characterization of the patients and poor integration of individual datasets. It attempts to establish and examine the relationship between tumor genomic characteristics and their radiological appearance.
Today, there is no dearth of well-organized and publicly available data repositories containing molecular data. On one hand, routine imaging data, unlike molecular phenotype data, is readily available in large quantities in patient records and can be relatively easily and cheaply collected retrospectively by investigators working at large clinical institutions. But on the other hand, while collecting patient outcomes data is relatively easy, in order to determine clinical significance, very long follow up periods are required. This means that usable outcomes data might not be available.
Radiogenomics bridges the gap by leveraging imperfect data along with prior knowledge of the relationship between outcomes with either imaging or genomics to draw new conclusions.
One example of this approach was taken for a study  by researchers to clear renal cell carcinoma for which a number of imaging features were correlated to mutations in VHS, PBRM1, SETD2, KDM5C, and BAP1 genes that were previously indicative of clinically significant factors of advanced grade, stage, and diminished survival prognosis. Thus, correlations were found between some of the imaging features and the gene mutations and through this discovery, imaging features were identified that are potentially predictive of outcomes.
Finally, beyond filling the gaps in knowledge, radiogenomics discoveries have a more basic significance of building a better understanding of the imaging representations of various molecular phenotypes, uncovering biological processes that are underlying phenotypes seen in imaging (i.e. causal relationship between the two), which could drive future discoveries in cancer research.
The main limitation of radiogenomics is related to the fact that if strong imaging and outcomes data are available, and the prediction of outcomes is the primary goal, the radiogenomic analysis may not bring an additional contribution to the analysis.
In radiogenomic analysis, investigators assess the association between imaging and genomic data. In order to conduct the analysis, specific features have to be extracted from the images. The imaging features are typically extracted by human readers, sometimes with some minor assistance from a computer program. Manual analysis of images has some disadvantages, most significantly, interobserver variability. Related to interobserver variability is the lack of precision when human observers conduct quantitative assessment of features such as mass volume, or volume of a specific part of a tumor.
Automatic feature extraction alleviates some of the issues through computer vision algorithms which are used to automatically segment the abnormality and extract a variety of features.
One such feature is the Predible Liver which enables the radiologist to don the researchers hat in the confines of daily practice by creating a structured quantified dataset of lesion characteristics in a rapid manner. Sign up for Beta!
- Dr. Mazurowski, Maciej, Radiogenomics: what it is and why it is important Journal of American College of Radiology
- Karlo CA, Di Paolo PL, Chaim J, Hakimi AA, Ostrovnaya I, Russo P, et al. Radiogenomics of clear cell renal cell carcinoma: associations between CT imaging features and mutations. Radiology. 2014; 270(2):464–471.