PhD Scientific Days 2018

Budapest, April 19–20, 2018

Computed Tomography Radiomics Can Identify Invasive and Nuclear Imaging Markers of Plaque Vulnerability

Kolossváry, Márton

Márton Kolossváry
MTA-SE Cardiovascular Imaging Research Group

Language of the presentation

English

Text of the abstract

Introduction:
Several morphologic and metabolic adverse plaque characteristics have been described by intravascular ultrasound (IVUS), optical coherence tomography (OCT) and NaF18-Positron Emission Tomography (NaF18-PET). It would be desirable to identify coronary plaques showing morphologic and/or metabolic vulnerability with a widely available non-invasive imaging modality. Coronary CT radiomics is a promising new technique which extracts quantitative metrics describing the heterogeneity and spatial complexity of lesions to create big-data datasets, where each abnormality is characterized by hundreds of different parameters. However, the discriminatory power of coronary CT radiomics to identify morphologic and/or metabolic plaque vulnerability is unknown.

Aims:
To assess the diagnostic accuracy of coronary computed tomography angiography (CTA) derived radiomic features to identify morphologic and metabolic plaque vulnerability, as compared to conventional qualitative and quantitative CT metrics.

Methods:
We analyzed 44 plaques in 25 patients using IVUS, OCT, NaF18-PET and coronary CTA. We assessed 7 conventional qualitative and quantitative plaque characteristics and calculated 935 radiomic parameters. Morphologic vulnerability was defined as a plaque with positive remodeling and posterior attenuation on IVUS and thin-cap fibroatheroma or microvessels or macrophage infiltration on OCT. Metabolic vulnerability was defined as NaF18 uptake >25% as compared to a reference lesion. We calculated receiver operating characteristics area under the curve (AUC) values using a 5-fold cross validation with 1000 repeats.

Results:
The best conventional CT metric’s AUC value was close to random (AUC=0.52), while the best radiomic feature had good diagnostic accuracy (AUC=0.74) to identify morphologic plaque vulnerability. Similarly, conventional CT metrics yielded a lower diagnostic accuracy than radiomics to identify metabolic vulnerability (AUC: 0.65 vs. 0.87; respectively).

Conclusions:
Coronary CTA radiomics outperforms conventional CT metrics to identify both morphologic and metabolic plaque vulnerability. Coronary CTA radiomics may allow robust identification morphologic and metabolic characteristics of plaque vulnerability currently only detectable invasively or with nuclear imaging techniques.

Data of the presenter

Doctoral School: Basic and Translational Medicine
Program: Cardiovascular Disorders: Physiology and Medicine of Ischaemic Circulatory Diseases
Supervisor: Pál Maurovich-Horvat
E-mail address: marton.kolossvary@cirg.hu