Automated peroperative assessment of stents apposition from OCT pullbacks

Comput Biol Med. 2015 Apr:59:98-105. doi: 10.1016/j.compbiomed.2014.12.012. Epub 2015 Jan 6.


This study's aim was to control the stents apposition by automatically analyzing endovascular optical coherence tomography (OCT) sequences. Lumen is detected using threshold, morphological and gradient operators to run a Dijkstra algorithm. Wrong detection tagged by the user and caused by bifurcation, struts'presence, thrombotic lesions or dissections can be corrected using a morphing algorithm. Struts are also segmented by computing symmetrical and morphological operators. Euclidian distance between detected struts and wall artery initializes a stent's complete distance map and missing data are interpolated with thin-plate spline functions. Rejection of detected outliers, regularization of parameters by generalized cross-validation and using the one-side cyclic property of the map also optimize accuracy. Several indices computed from the map provide quantitative values of malapposition. Algorithm was run on four in-vivo OCT sequences including different incomplete stent apposition's cases. Comparison with manual expert measurements validates the segmentation׳s accuracy and shows an almost perfect concordance of automated results.

Keywords: Biomedical image processing; Coronary artery stenting; Optical coherence tomography (OCT); Thin-plate spline (TPS).

MeSH terms

  • Algorithms
  • Endovascular Procedures / methods*
  • Humans
  • Image Processing, Computer-Assisted / methods*
  • Stents*
  • Surgery, Computer-Assisted
  • Tomography, Optical Coherence / methods*