The accuracy and interobserver variability in the assessment of coronary atherosclerotic plaques by optical frequency domain imaging: involving five observers with different levels of coronary imaging

Abstract

Abstract Background Whether optical frequency domain imaging (OFDI) images can realize pathological diagnosis of coronary atherosclerotic plaques, and whether its diagnostic accuracy of lesion types varies depending on the personal experience of the clinician caring for coronary intervention have not been elucidated. Purpose This study investigated the interobserver variability in characterizing atherosclerotic plaque types by OFDI for multiple OFDI observers with levels of different experience. Methods Three-hundred-thirty-three histological cross-sections from 21 autopsy hearts were co-registered with the corresponding OFDI images. Histological cross-sections were classified into the following 7 lesion types according to the modified AHA atherosclerosis classification by a single experienced pathologist blinded for OCT findings: adaptive intimal thickening (AIT), intimal xanthoma (IX), pathological intimal thickening (PIT), fibrous cap atheroma (FA), fibrocalcific plaque (FC), calcified nodule (CN), and healed erosion/rupture (HER). The five OFDI observers, unaware of the histological diagnosis, provided a single diagnosis for each corresponding OFDI image. The OFDI observer 1 was an expert interventional cardiologist with sufficient experience in OFDI imaging, followed by the OFDI observer 2, 3, and 4 as middle career interventional cardiologists who had completed training ten, seven, and four years. The OFDI observer 5 was a young career interventional cardiologist. The diagnostic accuracy of lesion types for each OFDI observer was determined taking histology as a gold standard. Results On histological analysis, 13% of histological cross-sections were diagnosed as AIT, 5% as IX, 23% as PIT, 25% as FA, 27% as FC, 2% as CN, and 5% as HER. The overall agreement between OFDI diagnosis and histopathologic diagnosis for OFDI observer 1 to 5 was 77%, 62%, 61%, 56%, and 46% (k values of 0.71, 0.54, 0.54, 0.45, and 0.33), respectively. Although the performance for characterizing AIT and FC was excellent and comparable among all OFDI observers, the sensitivity and positive predictive value for characterizing IX, PIT and FA varied depending on the OFDI observers' years of experience (Table). The main causes of false-positive or -negative diagnosis of FA were IX and PIT for all OFDI observers. Conclusion The diagnostic accuracy of atherosclerotic tissue properties from OFDI images correlated with the observers' years of experience, subspecialty training in coronary imaging, which suggests that the interpretation of OFDI images requires expertise and can be challenging to a less experienced reader. Table 1 Funding Acknowledgement Type of funding source: None