Histopathological analysis is the gold standard tool in the diagnosis of cancer. The most common way to perform this analysis consist of the extraction of tissue samples to examine it more closely under the microscope. Usually, tissue samples are extracted by either needle biopsy or surgery. However, this procedure is usually painful, risky and a time-consuming task. Moreover, in some cases tissue extraction may be insufficient and the extraction must be repeated. On the other hand, as part of cancer diagnosis the radiological examination is performed to know the location, size and shape of suspicious masses or tumors. Radiological imaging can capture information from the entire tumor, e.g., size, shape, location or even tissue heterogeneity, whereas histopathological images can capture local information of the tumoral tissue. Although histological and radiological analysis composes the basis of cancer diagnosis, the prognosis, treatment and follow-up of the disease are driven by a separate evaluation that is complementary in the best case.

In recent years, the computer-aided diagnosis (CAD) tools may monitor the progression of oncological diseases and their response to therapy becoming as an aid-tool to support clinical decisions. Radiomics is a recent and promising approach that is considered part of CAD tools. It aims to extract quantitative features from medical images, specifically radiological images, by using a combination of statistical analysis and machine learning methods. These features, termed radiomic features, have the potential to uncover disease characteristics that are difficult to identify for the human eye. Besides, the extraction and analysis of quantitative features from different radiological images such as magnetic resonance (MR), or computed tomography (CT) combined with either descriptive or predictive statistical models, have the potential to find out non-invasive biomarkers to support and improve clinical decisions.

Recently, radiomics has gained attention, in medical practice, specifically in oncology, since it could provide quantitative information of tumor phenotype by analyzing the relationship and distribution of pixels in 2D images or voxels in 3D volumes in radiological imaging. Therefore, extraction of radiomic features from tumor in radiological images could be considered as a kind of “digital biopsy” that could reduce the procedure time and costs associated with patient attention so as to extract a significative tissue sample. However, radiomics tackle some challenges before to use in the clinical practice. These challenges include the sensitivity to variations in image acquisition and reconstruction parameters, the reproducibility of the quantitative features and the explainability of the models, and other things. Nevertheless, Radiomics have the potential to allow a personalized treatments for cancer patients.

Sources:

Aerts, H. J., Velazquez, E. R., Leijenaar, R. T., Parmar, C., Grossmann, P., Carvalho, S., … & Lambin, P. (2014). Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nature communications, 5(1), 1-9.

Kumar, V., Gu, Y., Basu, S., Berglund, A., Eschrich, S. A., Schabath, M. B., … & Gillies, R. J. (2012). Radiomics: the process and the challenges. Magnetic resonance imaging, 30(9), 1234-1248.

Fornacon-Wood, I., Faivre-Finn, C., O’Connor, J. P., & Price, G. J. (2020). Radiomics as a personalized medicine tool in lung cancer: Separating the hope from the hype. Lung Cancer, 146, 197-208.