Machine-learning assisted detection of cancerous cell nuclei based on confocal imaging and fluorescent probes in hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is the most common primary liver cancer worldwide [1]. Liver biopsy and histopathology are crucial for the identification of tissue and cellular alterations, and diagnosis and staging of HCC [2]. By histological definition, there are macroscopic types and microscopic types that define the cellular subtypes of HCC. Interestingly, as well as in other cancers, also in HCC, the architecture of the nucleus is one of the elements that can differentiate neoplastic hepatocytes from healthy hepatocytes. In fact, a recent study has reported an association between prognosis and nuclear morphometric differences between HCC and adjacent tissue samples [3,4]. Given this prominent role of nuclear structure changes in diseased cells, several machine learning (ML) techniques have been developed based on quantitative information about the size and shape of a cell nucleus as well as the nucleus-cytoplasm ratio and chromatin consistency. In this regard, in breast cancer was showed “how” it is possible to correlate changes in heterochromatin with euchromatin ratios in normal and tumour cell lines, so as to recognize possible tumour cells by using artificial intelligence approaches (AI) [5]. However, a still missing emerging aspect of histological analysis, is the rapid identification by AI of scattered diseased cells for both HCC and tumours in general.