Abstract 15419: Automated Assessment of High-Risk Coronary Plaques Using Fully Integrated Catheter-Based Label-Free Fluorescence Lifetime Imaging (FLIm)-Optical Coherence Tomography (OCT)

IntroductionMachine learning (ML) can be applied to enhance biomedical imaging diagnostics, however, its feasibility in intracoronary multi-modal imaging has not been evaluated. This study aimed to investigate whether a fully-integrated FLIm-OCT incorporating ML into imaging analysis has the capabil...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Circulation (New York, N.Y.) N.Y.), 2018-11, Vol.138 (Suppl_1 Suppl 1), p.A15419-A15419
Hauptverfasser: Kim, Sunwon, Nam, Hyeong Soo, Kang, Woo Jae, Song, Joon Woo, Lee, Min Woo, Joo, Young Dae, Oh, Wang-Yuhl, Yoo, Hongki, Kim, Jin Won
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:IntroductionMachine learning (ML) can be applied to enhance biomedical imaging diagnostics, however, its feasibility in intracoronary multi-modal imaging has not been evaluated. This study aimed to investigate whether a fully-integrated FLIm-OCT incorporating ML into imaging analysis has the capability to identify the coronary plaque with high-risk biological features.Methods and resultsWe constructed a fully integrated, high-speed, multispectral FLIm-OCT imaging system to simultaneously visualize morphology and biochemical composition of atherosclerotic plaque. Rapid intracoronary imaging (pullback speed20 mm/s, rotation100 rps) under dextran flushing was safely performed in beating coronary arteries of atherosclerotic swine models. Along with detailed coronary microstructure by OCT, our multispectral FLIm could accurately visualize lifetime signature of key biochemical components of high-risk plaque in vivo (lipid, macrophage, and fibrous tissue). Lifetime values were different between lipid-rich inflamed vs. fibrotic plaque, p < 0.001). Intriguingly, significant differences were noted between macrophage-rich vs. lipid-rich regions (p < 0.0001, Figure), which were indistinguishable with current standalone OCT. Ex vivo fluorescence lifetime microscopy and immunostaining well corroborated the findings in vivo. Our ML classification algorithm, trained with the multiple sets of FLIm-OCT data and corresponding immunostaining sections, showed performance comparable to that of OCT imaging experts in classifying fibrotic vs. high-risk plaque. Furthermore, our framework demonstrated the capability to differentiate lipid vs. macrophage vs. mixed lesion (Figure).ConclusionsOur dual-modal, label-free FLIm-OCT with ML-based systematic signal analysis could provide high-resolution plaque imaging with further biochemical characterization. This highly translatable imaging strategy could offer new opportunity for detection of high-risk plaques in coronary artery.
ISSN:0009-7322
1524-4539