Executive Summary : | Cancer is a leading cause of death worldwide, accounting for nearly 10 million deaths in 2020 alone. As the incidence rate of cancer and its mortality rate is rapidly increasing, different modalities such as modern laparoscopy surgery, robotic surgery, tumor adjuvant therapy, and other new technologies are increasingly being adopted to prolong survival and reduce local recurrences. Importantly, the epidemiologic trend of cancer prevalence and mortality is disproportionately distributed with an increase in low- and middle-income countries (LMICs) than in high-income countries (HICs) and it warrants particular concern. This is evident in the case of head and neck cancers (HNC) for example which account for 25%–30% of all cancer in India compared to 1%–4% in HICs. It is estimated that over 200,000 cases of HNC occur each year in the country, most of which are diagnosed at an advanced stage which significantly reduces the survival rate, even after curative treatment. The intra-tumor heterogeneity (ITH) and histopathology of HNC also affect the clinicians’ decisions in providing the best treatment options that can mitigate the disease progression. Therefore, to improve the patient’s survival rate, early diagnosis and prognosis prediction of HNC are of utmost importance and are currently the need of the hour. Medical imaging is an integral part of cancer clinical protocols and it provides morphological, structural, metabolic, and functional information. HNC is highly dependent on imaging, including computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) besides tissue and blood tests, for the diagnosis, treatment plan, image-guided interventions, and response assessment. Imaging data captured from these imaging modalities are routinely acquired, exponentially increasing in their volume, variety, and velocity. Curating such large data, however, highly depends on the experience of the radiologist and can be extremely time-consuming, therefore, impractical for routine implementation in the clinical setting. To counter this problem, the extraction of minable information from images can be explored via explicit (handcrafted/engineered) and deep radiomics frameworks. Radiomics was first coined by Lambin et al. and is an emerging field in medical imaging used in the diagnostics and response prediction of various forms of cancer including HNC. It is used in the characterization of “tumor phenotypes such as size (volume), shape (sphericity, compactness), texture (voxel heterogeneity, coarseness, contrast) and tumoral intensity (uniformity, entropy)”, through high throughput extraction of quantitative features from routine radiological images. Radiomics is computationally attractive due to its scalability, efficiency, and precision. |