Validation of a semi-automatic software for optical coherence tomography - analysis in heart transplanted patients

Abstract

Optical Coherence Tomography (OCT) is an intravascular imaging modality enabling detailed evaluation of cardiac allograft vasculopathy (CAV) after heart transplantation (HTx). However, its clinical application remains hampered by time-consuming manual quantitative analysis. We aimed to validate a semi-automated quantitative OCT analysis software (Iowa Coronary Wall Analyzer, ICWA-OCT) to improve OCT-analysis in HTx patients. 23 patients underwent OCT evaluation of all three major coronary arteries at 3 months (3M) and 12 months (12M) after HTx. We analyzed OCT recordings using the semiautomatic software and compared results with measurements from a validated manual software. For semi-automated analysis, 31,228 frames from 114 vessels were available. The validation was based on a subset of 4287 matched frames. We applied mixed model statistics to accommodate the multilevel data structure with method as a fixed effect. Lumen (minimum, mean, maximum) and media (mean, maximum) metrics showed no significant differences. Mean and maximum intima area were underestimated by the semi-automated method (β-method_mean = - 0.289 mm², p < 0.01; β-method_max = - 0.695 mm², p < 0.01). Bland-Altman analyses showed increasing semi-automatic underestimation of intima measurements with increasing intimal extent. Comparing 3M to 12M progression between methods, mean intimal area showed minor underestimation (β-method_mean = - 1.03 mm², p = 0.04). Lumen and media metrics showed excellent agreement between the manual and semi-automated method. Intima metrics and progressions from 3M to 12M were slightly underestimated by the semi-automated OCT software with unknown clinical relevance. The semi-automated software has the future potential to provide robust and time-saving evaluation of CAV progression.

Keywords: Cardiac allograft vasculopathy; Heart transplantation; Intravascular imaging; Optical coherence tomography; Validation.

Fig. 1

Iowa Coronary Wall Analyzer for OCT (ICWA-OCT). a Enlarged cross-sectional view of non-edited frame. b Enlarged cross-sectional view of vessel layer segmentation suggested by the semi-automatic software. Lumen is found within the red line, intima layer between red and purple lines and media layer between purple and green lines. c Cross-sectional view of angular analysis performed by the semi-automatic software. Included angles are displayed by the outmost red dotted circle (angular analysis)

Fig. 2

Scatter and Bland Altman plots of mean intima area, maximum intima area, minimum lumen area and maximum intima thickness estimated by the manual and semi-automatic method (manual as reference). Mixed model equations are presented with 95%-confidence intervals (CI) and mean differences of Bland Altman plots are presented with 95%-Limits of Agreement (LoA)

Fig. 3

Change from 3 to 12M (delta parameters) for mean intima area, maximum intima area and minimum lumen area. Depiction in scatter and Bland–Altman plots comparing manual and semi-automated analyses using matched frames (manual as reference)

Fig. 4

Comparison of cases with different degrees of CAV by angiography and OCT recordings with manual and semi-automatic methods. All images are from 12M examinations. Segmentation borders for OCT-analyses: Red, lumen; purple, internal elastic membrane; green, external elastic membrane

Fig. 5

Boxplots showing maximum intima area estimated by semi-automatic and manual method for each vessel at 3 months (3M) and 12 months (12M) using matched frames. Median and IQR provided