Keywords:
CAD, Physics, CT, Computer applications, Computer Applications-Detection, diagnosis, Tissue characterisation
Authors:
G. Ficarra1, E. Barabino2, M. Verda3, S. Casella4, S. Caprioli5, G. Cittadini1; 1Genoa/IT, 2Genova/IT, 3Imperia/IT, 4Savona/IT, 5Arenzano/IT
DOI:
10.26044/ecr2019/C-2626
Aims and objectives
Current medical practice involves extensive use of non-invasive imaging techniques,
such as magnetic resonance (MR),
computed tomography (CT),
and metabolic (PET) imaging,
especially among the oncologic patients. Several studies have hypothesized that tumour characteristics at the cellular and genetic levels are reflected in the phenotypic patterns that can be captured with imaging 1,2,3;despite this promise,
tumour features are often assessed visually and described qualitatively by imaging specialists and subjective assessment suffers from large intra and inter-observer variability 4,5,6.
Nevertheless even biological samples fail to represent the whole complexity of a lesion,
evaluating just a small,
usually random sampled part of the tumor; therefore imaging tools with potential to characterize the complexity of the neoplasm would be of tremendous value. Although it is not a new tool7, renewed interest in computed tomographic (CT) texture analysis (CTTA) arised in recent years in the attempt to find a useful biomarker that allows assessment and quantification of tumor spatial heterogeneity. CTTA is just a tool in the growing field of radiomics, which strives to develop and use quantitative methods (quantitative image features) to characterize phenotypic differences of cancers and/or cancer subtypes,
and to do so through images that can be acquired during routine medical practice.
Even if CTTA has shown promising results,
the conclusions reached have to be taken carefully because results can vary greatly; many studies have analyzed the influence of CT acquisition and reconstruction parameters on CTTA derived features 8,9,10,11. However,
even the use of the same acquisition protocol for different CT scanners does not ensure reproducibility.
This is not only because of differences in hardware components between vendors,
but also because of anatomic,
physiologic,
and positional variations.
In the present study we aimed to investigate the impact of CT acquisition parameters on TA features; furthermore we aimed to understand which features are more susceptible to variation in acquisition parameters.