Identification of Neoatherosclerosis in ISR Patients Based on Artifical Intelligence

Overview

Based on the large population of patients, in-stent restenosis (ISR) is still an important problem in the field of cardiovascular disease. How to reduce the incidence of ISR and the treatment of ISR has become the focus and hot spot. The 2018 ESC Guidelines for Cardiovascular Intervention recommends treatment of ISR under the guidance of intravascular ultrasound (IVUS), or optical coherent tomography (OCT). Circulation published a new Waksman ISR classification based on mechanisms and components of the restenosis tissue, which provides guidance for treatment strategy. Because of its good resolution, OCT makes it more accurate to distinguish the components of vascular tissue, thus providing a decision-making basis for interventional therapy. OCT examination can obtain the characteristics of the ISR more precisely. Neoatherosclerosis (NA), is one of the ISR types and accounts for more stent failure and target lesion failure than other types. Identification NA is important for decision-making of interventional therapy. However, the acquisition and analysis of OCT images not only need the digital angiography machine (DSA) equipped with the majority of hospitals, but also need professional OCT imaging equipment and technicians. Patients with severely CKD cannot bear OCT examination because of the large amount of contrast agent. OCT catheter is more than ten times the price of the CAG catheter. Therefore, identification of NA by the use of artificial intelligence (AI) is of significance to set therapeutic strategy for ISR patients, especially in patients with CKD. Our study retrospectively analyzed CAG images and OCT images of ISR patients obtained from Jan 1st,2015 to Oct 31st,2020. Identify NA by analyzing OCT images, build up U-net and V-net to analyze the CAG and OCT images, and finally build up an identification system of NA based on CAG images by AI. This study has been approved by Ethics Committee of Chinese PLA General Hospital (S2018-033-01)

Drug Eluting Stents (DES) reduce the rate of in-stent restenosis (ISR) to 3.6-10%. Based on the large population of patients, ISR is still an important problem in the field of cardiovascular disease. How to reduce the incidence of ISR and the treatment of ISR has become the focus and hot spot. The 2018 ESC Guidelines for Cardiovascular Intervention recommends treatment of ISR under the guidance of intravascular ultrasound (IVUS), or optical coherent tomography (OCT). The European Expert Consensus on Intravascular Imaging, published in 2018, recommends finding the underlying mechanisms of ISR through intravascular imaging guidance (IVUS or OCT), and determining therapeutic strategies based on the mechanisms. Circulation published a new Waksman ISR classification based on mechanisms and components of the restenosis tissue, which provides guidance of treatment strategy. The use of intravascular imaging to identify and classify the types and mechanisms is very important for ISR treatment strategy. Because of its good resolution, OCT makes it more accurate to distinguish the components of vascular tissue, thus providing a decision-making basis for interventional therapy. OCT examination can obtain the characteristics of ISR more precisely. Neoatherosclerosis (NA), is one of the ISR types and accounts for more stent failure and target lesion failure than other types. Identification of NA is important for decision-making of interventional therapy. However, the acquisition and analysis of OCT images not only need the digital angiography machine (DSA) equipped with the majority of hospitals, but also need professional OCT imaging equipment and technicians. Patients with severely CKD cannot bear OCT examination because of the large amount of contrast agent. OCT catheter is more than ten times the price of the CAG catheter. Therefore, identification of NA by the use of artificial intelligence (AI) is of significance to set therapeutic strategy for ISR patients, especially in patients with CKD. Our study retrospectively analyzed CAG images and OCT images of ISR patients obtained from Jan 1st,2015 to Jan 31st,2020. Offline OCT analysis was performed using dedicated software (Light Lab Imaging Inc, Westford, MA). All images were analyzed at every frame in the stents by 2 independent investigators, who were blinded to the angiographic and clinical findings. Identify NA by analyzing OCT images, build up U-net and V-net to analyze the CAG and OCT images, and finally build up an identification system of NA based on CAG images by AI. This study has been approved by Ethics Committee of Chinese PLA General Hospital (S2018-033-01)