From immunohistochemistry to machine learning-based patient stratification by tumour proliferation characteristics in...

It is unanimously acknowledged that head and neck carcinomas (HNC) are a group of cancer with high proliferative ability, leading to aggressive behaviour and poor response to treatment. Of the arsenal of biomarkers representative for cellular proliferation, the Ki-67 nuclear protein which is present in active cells was identified as one of the most accurate to assist in outcome prediction based on proliferation rate and implicitly, in patient stratification. While immunohistochemistry is a traditional method for Ki-67 quantification, developments in the field of radiomics encouraged the use of machine learning algorithms on patients' images to identify in a non-invasive and cost-effective manner, specific imaging patterns correlated with Ki-67 expression levels. The aim of the current work was to describe the current status of proliferation markers specific to head and neck cancers with emphasis on Ki-67, and to ascertain the latest research on machine learning studies that identified radiomic signatures pertaining to proliferation to be used for prediction of Ki-67 expression levels. While most of these radiomic studies are newly published and limited in number, the results are supportive of the role of machine learning in clinical outcome predictions based on proliferation rate and warrant further investigations in the field.