Automated accurate lumen segmentation using L-mode interpolation for three-dimensional intravascular optical coherence tomography.

Intravascular optical coherence tomography (IVOCT) lumen-based computational flow dynamics (CFD) enables physiologic evaluations such as of the fractional flow reserve (FFR) and wall sheer stress. In this study, we developed an accurate, time-efficient method for extracting lumen contours of the coronary artery. The contours of cross-sectional images containing wide intimal discontinuities due to guide wire shadowing and large bifurcations were delineated by utilizing the natural longitudinal lumen continuity of the arteries. Our algorithm was applied to 5931 pre-intervention OCT images acquired from 40 patients. For a quantitative comparison, the images were also processed through manual segmentation (the reference standard) and automated ones utilizing cross-sectional and longitudinal continuities. The results showed that the proposed algorithm outperforms other schemes, exhibiting a strong correlation (R = 0.988) and overlapping and non-overlapping area ratios of 0.931 and 0.101, respectively. To examine the accuracy of the OCT-derived FFR calculated using the proposed scheme, a CFD simulation of a three-dimensional coronary geometry was performed. The strong correlation with a manual lumen-derived FFR (R = 0.978) further demonstrated the reliability and accuracy of our algorithm with potential applications in clinical settings.

Regression-based technique for improved optical rangefinding using tunable focus lenses.

In this paper, we present a novel method of target range estimation by tuning the spot size of a Gaussian beam at the plane of a reflective target. The beam spot size tuning is achieved through the use of a tunable focus lens (TFL). Using a carefully aligned sensor assembly, the diameter of the reflected beam is recorded at the plane of an imaging detector for different TFL focal length settings. This dataset is then used to estimate the distance of the target from the TFL. The proposed rangefinder is compact and requires minimal post-data-acquisition signal processing resulting in a fast response time compared to other spatial signal processing-based sensor designs. The estimation of target distance through a multiple data-point measurement dataset also ensures that the proposed method is robust to errors associated with obtaining range estimates from a single measurement data point. Experimental results demonstrate an excellent agreement with theory. With our proposed estimation method, we show a significant improvement in the measurement dynamic range of the sensor as well as its resolution compared to similar sensing schemes in prior art. We also experimentally demonstrate the possibility to extend the measurement dynamic range by incorporating a bias lens of a fixed focal length with the sensor module. The proposed sensor module is electronically controlled and consequently can be fully automated and compactly packaged with the use of commercially available miniature optical components.