Development and Validation of Deep Learning Model for Intravascular Ultrasound Image Segmentation

Abstract

Aims. This study was conducted to develop and validate a deep learning model for delineating intravascular ultrasound (IVUS) images of coronary arteries. Materials and Methods. Using atotal of 1240 40-MHz IVUS pullbacks with 191,407 frames, the model for lumen and external elastic membrane (EEM) segmentation was developed. Both frame- and vessel-level performances and clinical impact of the model on 3-year cardiovascular events were evaluated in the independent data sets. Results. In the test set, the Dice similarity coefficients (DSC) were 0.966 ± 0.025 and 0.982 ± 0.017 for the lumen and EEM, respectively. Even at sites of extensive attenuation, the frame-level performance was excellent (DSCs > 0.96 for the lumen and EEM). The model (vs. the expert) showed a better temporal consistency for contouring the EEM. The agreement between the model- vs. the expert-derived cross-sectional and volumetric measurements was good in the independent retrospective cohort (all, intra-class coefficients > 0.94). The model-derived PAV>52.5% (area under curve 0.70, sensitivity 71% and specificity 67%) and plaque burden at the MLA site (area under curve 0.72, sensitivity 72% and specificity 66%) best predicted 3-year cardiac death and nonculprit-related target vessel revascularization, respectively. In the stented segment, the DSCs > 0.96 for contouring lumen and EEM were achieved by fine-tuning. Applied to the 60-MHz IVUS images, the DSCs were > 0.97. In the external cohort with 45-MHz IVUS, the DSCs were > 0.96. Conclusion. The deep learning model accurately delineated vascular geometry, which may be cost-saving and support clinical decision-making.