116 - The Hypertensive Heart: Can Radiomics Assess Micro-anatomical Changes Relevant to Early Diagnosis?

Poster Board Number: 116
There are separate poster presentation times for odd and even posters.
Odd poster #s – first hour
Even poster #s – second hour

Abstract Body : Hypertension or high blood pressure is a common factor for the development of life-threatening cardiovascular diseases. Hypertension is known to induce sub-clinical changes in heart geometry and microarchitecture that conventional methods are inadequate at detecting. Radiomics is an emerging field and has the potential to automate the detection of subclinical changes in the phenotype of the hypertensive heart in the early stages of disease modification by combining micro-anatomy feature analysis and machine learning of cardiac imaging to quantitatively analyse risk factors correlating with the hypertensive heart. The challenge for this emerging field is the lack of standardised methodology for feature identification and analysis. This study therefore aimed to systematic analyse of radiomic studies on the hypertensive heart to standardise the features significant across a range of imaging modalities, and to apply this to a test cardiac MRI dataset, and to propose an optimised methodology to standardising data collection.
A comprehensive systematic review was undertaken using the Cochrane, PUBMED, Scopus, and EMBASE databases. The search strategy included all research reports up to 27 May 2022. A total of 2160 reports was initially retrieved and following screening yielded 34 studies for analysis.
Features were analysed on a range of imaging modalities in the studies: MRI (n=16); CT(n=14); echocardiograms(n=2); ultrasound(n=1); and echocardiograms and CT angiography (n=1). While extracted features across the studies ranging from 28 to 4,440 features, for hypertensive hearts. second-order features were most common across studies, indicating the significance of micro-changes to the heart architecture. Less frequently reported significant features were shape (n = 9) and higher-order features (n = 5), respectively. Radiomic features related to size category were deemed significant less frequent, with only one radiomic feature (volume surface area) found to be significant. Of the 34 studies, 13 studies reported left ventricle geometry, and 5 atherosclerotic plaques.
Radiomic modelling of the hypertensive heart is feasible and more accurate over conventional quantitative parameters, able to discern changes currently not identifiable. Standardising methodology relevant to reliability of feature identification and reproducibility is urgently required to inform informatics and systems development.