Exposing Cancer's Complexity Using Radiomics in Clinical Imaging An Investigation on the Role of Histogram Analysis as Imaging Biomarker to Unravel Intra-Tumour Heterogeneity

Thanks to the most advanced investigation techniques, cancer is showing to be something more complex than we ever imagined. Genomic pattern, epigenetic modifications, environmental and life-style influences leads to subjective expression of the disease. In addition, cancer can be extremely heterogeneous intrinsically, and does not stand still but changes over time. These hallmarks can explain how cancer adapts to therapies, evolving to something than can be totally different from the beginning of the disease. It's an expression of Darwin evolution. Spatial heterogeneity can be found among different tumors and within each lesion, which manifests at genomic, phenotypic, and physiologic levels. Today we know that heterogeneity is a hallmark of malignant tumors. Usually intratumor heterogeneity tends to increase as tumors grow and may increase or decrease following the response to the therapy. This means that tumor heterogeneity must be explored as prognostic tool, but how do we measure this heterogeneity? Clinical imaging allows to quantify this heterogeneity thanks to Radiomics, which extracts quantitative features from images (especially from computed tomography [CT], magnetic resonance [MR], and positron emission tomography [PET] images). The link of these imaging parameters to different phenotypes or genotypes enables the mapping of biologic heterogeneity of tumors, from which inference on gene expression, signaling pathway activity, and tumor microenvironment features can be obtained. These features have the potentiality to become a powerful tool to unravel tumor, providing quantitative information that allows a better phenotypization. In this work we want to show how a subset of radiomic features connected to histogram analysis, in particular skewness, kurtosis and Shannon entropy, evaluated in images of patients with different kinds of cancer, show diagnostic power to differentiate healthy from ill tissues. We will conclude introducing problems linked to the lack of connection between the complexity estimated with radiomics and the underlying biological model.