Keywords:
Cancer, Experimental investigations, MR, Oncology, Head and neck, Hyperplasia / Hypertrophy, Tissue characterisation
Authors:
S. Caprioli1, S. Casella2, M. Verda3, G. Ficarra4, E. Barabino5, G. Cittadini4; 1Arenzano/IT, 2Savona/IT, 3Imperia/IT, 4Genoa/IT, 5Genova/IT
DOI:
10.26044/ecr2019/C-3352
Aims and objectives
Since its introduction in clinical practice,
MRI played a fundamental role in the diagnosis of Head and Neck cancer.
DWI is an MR technique that studies molecular diffusion,
in particular the Brownian motion of water protons in biologic tissue.
Neoplastic structures are more densely cellulated and have a larger number of cellular membranes; due to that,
they present a greater impediment to molecular diffusion and lower Apparent Diffusion Coefficient (ADC) values than normal structures; therefore,
DWI improves the detection of cancer especially in head and neck [1].
Oropharynx can represent a challenging district for the radiologist in MR imaging and the presence of hyperplastic lymphoid tissue may hide small foci of Squamous Cell Carcinoma (SCC).
Since radiologic diagnosis of Oropharyngeal SCC is primarily based on morphological changes in the base of the tongue and enlargement or asymmetry in palatine tonsils,
detection of SCC in patients with enlarged lymphoid tissue may be very challenging.
Moreover,
lymphoid tissue of palatine tonsil and tongue base has restricted diffusion,
as Oropharyngeal SCCs have.
Bhatia et al.
stated that SCCs have higher mean ADC values compared to Lymphatic Hyperplasia (LH).
This phenomenon could be referred to the loss of the typical architecture of lymphoid tissue,
that usually predisposes to dense cellularity and,
therefore,
to restricted diffusion,
and to the acquiring of a more heterogeneous and deconstructed environment that could possibly explain a higher ADC value [2].
Radiomics is a technique that allows to extract a large amount of data from medical images and a promising method in non-invasive tissue characterization.
Machine learning is a form of artificial intelligence that can create predictive models from large dataset and its synergic effect in classification when combined with radiomic data is well described in literature [3].
The aim of our study is to investigate the utility of Radiomics and Machine Learning in differentiating Oropharyngeal SCC from Lymphatic Hyperplasia.