Elsevier

Atherosclerosis

Volume 345, March 2022, Pages 15-25
Atherosclerosis

Machine learning for atherosclerotic tissue component classification in combined near-infrared spectroscopy intravascular ultrasound imaging: Validation against histology

https://doi.org/10.1016/j.atherosclerosis.2022.01.021Get rights and content

Highlights

  • IVUS grey-scale pixel intensity analysis has limited accuracy in assessing plaque composition.

  • NIRS-signal was combined with a machine learning classifier trained to distinguish tissue types using IVUS pixel intensities.

  • The combined approach improved plaque characterization accuracy and was superior to stand-alone IVUS.

  • This approach may be advantageous for vulnerable plaque detection and for monitoring plaque evolution.

Abstract

Background and aims

Accurate classification of plaque composition is essential for treatment planning. Intravascular ultrasound (IVUS) has limited efficacy in assessing tissue types, while near-infrared spectroscopy (NIRS) provides complementary information to IVUS but lacks depth information. The aim of this study is to train and assess the efficacy of a machine learning classifier for plaque component classification that relies on IVUS echogenicity and NIRS-signal, using histology as reference standard.

Methods

Matched NIRS-IVUS and histology images from 15 cadaveric human coronary arteries were analyzed (10 vessels were used for training and 5 for testing). Fibrous/pathological intimal thickening (F-PIT), early necrotic core (ENC), late necrotic core (LNC), and calcific tissue regions-of-interest were detected on histology and superimposed onto IVUS frames. The pixel intensities of these tissue types from the training set were used to train a J48 classifier for plaque characterization (ECHO-classification). To aid differentiation of F-PIT from necrotic cores, the NIRS-signal was used to classify non-calcific pixels outside yellow-spot regions as F-PIT (ECHO-NIRS classification). The performance of ECHO and ECHO-NIRS classifications were validated against histology.

Results

262 matched frames were included in the analysis (162 constituted the training set and 100 the test set). The pixel intensities of F-PIT and ENC were similar and thus these two tissues could not be differentiated by echogenicity. With ENC and LNC as a single class, ECHO-classification showed good agreement with histology for detecting calcific and F-PIT tissues but had poor efficacy for necrotic cores (recall 0.59 and precision 0.29). Similar results were found when F-PIT and ENC were treated as a single class (recall and precision for LNC 0.78 and 0.33, respectively). ECHO-NIRS classification improved necrotic core and LNC detection, resulting in an increase of the overall accuracy of both models, from 81.4% to 91.8%, and from 87.9% to 94.7%, respectively. Comparable performance of the two models was seen in the test set where the overall accuracy of ECHO-NIRS classification was 95.0% and 95.5%, respectively.

Conclusions

The combination of echogenicity with NIRS-signal appears capable of overcoming limitations of echogenicity, enabling more accurate characterization of plaque components.

Introduction

The assessment of coronary plaque composition is essential in treatment planning and predicting cardiovascular events. Calcific atherosclerotic plaques have been associated with a higher risk of procedural complications, stent underexpansion and suboptimal percutaneous coronary intervention (PCI) results [[1], [2]], [[,2] while lipid-rich lesions appear to be associated with non-reflow and peri-procedural myocardial infarction [[3], [4], [5]]. Furthermore, cumulative data have demonstrated that lipid-rich atherosclerotic plaques are vulnerable lesions and predict future adverse cardiovascular events [[6], [7], [8], [9]].

Over recent years several methodologies aiming to accurately characterize and quantify plaque components in intravascular imaging have been introduced. Echogenicity [10] and radiofrequency backscatter analysis [[11], [12], [13]] were the first approaches proposed for reliable detection of different tissue types in intravascular ultrasound (IVUS) while in optical coherence tomography (OCT), backscatter and attenuation coefficient analysis [[14], [15], [16]] of the reflected signal, as well as machine learning techniques [17,18] have been proposed. However, none of them have found broad application in current clinical practice due to a lack of widely available, user-friendly software incorporating these methodologies or robust and consistent histological evidence to support their accuracy.

To address this unmet need, near-infrared spectroscopy (NIRS)-IVUS has been introduced. NIRS-IVUS appears capable of overcoming the limitations of standalone intravascular imaging as NIRS can accurately identify necrotic core (NC) tissue, while IVUS can detect the presence of calcium [19]. However, while NIRS can estimate the circumferential arc of NC, it is unable to provide explicit depth information, differentiate superficial from deeply-embedded NC tissue and assess its area and volumetric burden. Similarly, there is no methodology today for the automated quantification of the calcific tissue in IVUS images.

The aim of this study is to investigate for the first time the efficacy of tissue echogenicity combined with NIRS in accurately detecting distribution of plaque components and quantifying their burden, using histology as reference standard.

Section snippets

Studied patients

We retrospectively analyzed NIRS-IVUS and histological data from excised cadaver hearts collected in a previously described study that aimed to examine the efficacy of NIRS imaging in detecting NC [International Institute for the Advancement of Medicine, Edison, NJ, and Asterand Bioscience (BioIVT), Detroit, MI] [20]. Donated hearts were received within 48 h of death, with intravascular imaging performed within 96 h. Totally occluded segments and those with minimum luminal diameter <1 mm were

Results

Fifteen coronary arteries from thirteen autopsied hearts were included in the analysis. Baseline heart donor demographics are shown in Supplementary Table 2 and baseline IVUS measurements in Supplementary Table 3. The mean lengths of the studied segments was 46.0 ± 12.5 mm for the training set and 51.3 ± 8.4 mm for the test set, providing 346 histological sections for matching (220 for the training set and 126 for the test set). From these, 42 histological sections were excluded because of

Discussion

In this study, we examined for the first time the efficacy of echogenicity combined with the information gleaned from NIRS in characterizing plaque composition in NIRS-IVUS images. We found that: 1) NIRS is very sensitive in detecting ENC and LNC tissue, 2) IVUS pixel intensity enables accurate detection of calcific ROIs but it is unable to differentiate F-PIT from ENC and has limited efficacy in detecting LNC ROIs, and 3) the information provided by NIRS improves the efficacy of echogenicity

CRediT authorship contribution statement

Retesh Bajaj: Methodology, Investigation, Data curation, Formal analysis, Writing – original draft, Validation. Jeroen Eggermont: Methodology, Software, Validation, Writing – review & editing, and. Stephanie J. Grainger: Resources, Data curation, Writing – review & editing. Lorenz Räber: Writing – review & editing. Ramya Parasa: Writing – review & editing. Ameer Hamid A. Khan: Writing – review & editing. Christos Costa: Writing – review & editing. Emrah Erdogan: Writing – review & editing.

Declaration of competing interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: MJH and SJG are employees of Infraredx. All other authors have no conflicts of interests to declare.

Acknowledgements

The authors wish to acknowledge the Cardiovascular Devices Hub at the Centre for Cardiovascular Medicine and Devices, Queen Mary University of London for supporting the present study. RB, RP, MA, AM, AB and CVB are funded by Barts NIHR Biomedical Research Centre, London, UK.

References (52)

  • Y. Shiono et al.

    Impact of attenuated plaque as detected by intravascular ultrasound on the occurrence of microvascular obstruction after percutaneous coronary intervention in patients with ST-segment elevation myocardial infarction, JACC

    Cardiovasc. Interv.

    (2013)
  • J. Pu et al.

    Insights into echo-attenuated plaques, echolucent plaques, and plaques with spotty calcification

    J. Am. Coll. Cardiol.

    (2014)
  • P.A. Calvert et al.

    Association between IVUS findings and adverse outcomes in patients with coronary artery disease: the VIVA (VH-IVUS in vulnerable atherosclerosis) study

    JACC (J. Am. Coll. Cardiol.): Cardiovasc. Imag.

    (2011)
  • M. Okubo et al.

    Development of integrated backscatter intravascular ultrasound for tissue characterization of coronary plaques

    Ultrasound Med. Biol.

    (2008)
  • O. Manfrini et al.

    Sources of error and interpretation of plaque morphology by optical coherence tomography

    Am. J. Cardiol.

    (2006)
  • R. Bajaj et al.

    Advanced deep learning methodology for accurate, real-time segmentation of high-resolution intravascular ultrasound images

    Int. J. Cardiol.

    (2021)
  • R. Hoffmann et al.

    Treatment of calcified coronary lesions with Palmaz-Schatz stents. An intravascular ultrasound study

    Eur. Heart J.

    (1998)
  • C.V. Bourantas et al.

    Prognostic implications of coronary calcification in patients with obstructive coronary artery disease treated by percutaneous coronary intervention: a patient-level pooled analysis of 7 contemporary stent trials

    Heart

    (2014)
  • Y.J. Hong et al.

    Impact of plaque components on no-reflow phenomenon after stent deployment in patients with acute coronary syndrome: a virtual histology-intravascular ultrasound analysis

    Eur. Heart J.

    (2011)
  • A. Tanaka et al.

    Lipid-rich plaque and myocardial perfusion after successful stenting in patients with non-ST-segment elevation acute coronary syndrome: an optical coherence tomography study

    Eur. Heart J.

    (2009)
  • F. Prati et al.

    Relationship between coronary plaque morphology of the left anterior descending artery and 12 months clinical outcome: the CLIMA study

    Eur. Heart J.

    (2020)
  • G.W. Stone et al.

    A prospective natural-history study of coronary atherosclerosis

    N. Engl. J. Med.

    (2011)
  • N. Bruining et al.

    Three-dimensional and quantitative analysis of atherosclerotic plaque composition by automated differential echogenicity

    Cathet. Cardiovasc. Interv.

    (2007)
  • A. Nair et al.

    Coronary plaque classification with intravascular ultrasound radiofrequency data analysis

    Circulation

    (2002)
  • A. Nair et al.

    Automated coronary plaque characterisation with intravascular ultrasound backscatter: ex vivo validation, EuroIntervention

    J. EuroPCR Collaboration Work. Group Interventional Cardiol. Eur. Soc. Cardiol.

    (2007)
  • M. Kawasaki et al.

    In vivo quantitative tissue characterization of human coronary arterial plaques by use of integrated backscatter intravascular ultrasound and comparison with angioscopic findings

    Circulation

    (2002)
  • View full text