Feature Of The Week: 6/26/09: Fully Automatic Three-Dimensional (3D) Quantitative Analysis of Intracoronary Optical Coherence Tomography: Method and Validation
Feature Of The Week: 6/26/09: Optical Coherence Tomography continues to make tremendous progress in the diagnosis and treatment of coronary artery disease and second generation systems have accelerated that progress. The Thoraxcenter has been a leading organizations in this field as can be seen by clicking the Cardiology tag an scrolling down to "Organizations in the News". Described below is some recent work by Nico Bruining et al which discusses fully automatic three-dimensional quantitative analysis of intracoronary OCT.
Coronary artery disease (CAD) is one of the most diagnosed pathologies in humans and still one of the leading causes of death. Although great progress has been made in treatment of CAD, for both acute interventions as well as pharmacological treatment, there is still much room for further improvements. Research in this area is heavily depending on coronary artery imaging and to study into detail coronary atherosclerosis mostly intracoronary imaging methods are applied. Until recently intracoronary ultrasound (ICUS) was the de-facto imaging method of choice for studies in which new interventional techniques (e.g. new stent designs) and/or pharmacological treatments (e.g. atherosclerosis progression-regression) were applied.
The great advantage of OCT over ICUS is its image resolution which is approximately 10 times higher than that of ICUS and close to that of histopathology. This allows studying details of the coronary artery which previously were not possible with ICUS such as by example detailed imaging of endotheliazation of individual stent struts.
Recently, enormous progress has been made in the optical coherence tomography (OCT) imaging for coronary arteries. The OCT technique has been implemented onto a tiny wire which can be easily introduced into the artery via a guiding catheter. The first generation OCT required a blockage of the artery (by inflating a non-dilatating balloon), and thus to a shortage of oxygen supply leading to discomfort (or even pain) for the patient. This must be performed to be able to remove by flush (with saline) the blood from the artery since blood blocks the light signal preventing imaging of the artery wall. However, the 2nd generation coronary OCT is capable of acquiring images at high speeds (>120 frames/s) allowing imaging the coronary artery in a few seconds which can be accomplished by a simple flush without the necessity to block the vessel. This makes OCT more clinical accepted.
To use an imaging method in a study, quantitative parameters derived from these images are mandatory as has been shown in the past by quantitative coronary angiography (QCA) and quantitative ICUS (QCU). As OCT has great similarities with ICUS (ICUS is based on sound and OCT on light) analysis methods could be very much similar. Preferably, most quantitative methods should be as much automated as possible to 1) reduce analysis time, which can be very long as pullbacks of ICUS and OCT can contain hundreds, or even thousands, of individual cross-sectional frames; and 2) to rule-out, as much as possible, observer related induced biases. For QCA that is possible and is in use for the past two decades, however, for QCU until today no full-automated contour detection method has been presented and most quantitative software methods are incorporating semi-automated methods. This is mostly due to the “fuzzy” appearance of the coronary artery wall in ICUS images caused in many cases by blood speckling. Currently most published QOCT approaches are actually QCU software methods adapted to import OCT images and the contour detection is mostly performed semi-automated or even manually.
However, since OCT images are not suffering from any interference of the blood, which is flushed away, and the “sharp” defined blood-lumen interface in the OCT images, it could maybe possible to detect these lumen contours fully-automated for QOCT. A study investigating the possibility for full-automated QOCT is now presented by a team from the Thoraxcenter of the Erasmus MC and the TU-Delft. This study presents a method to detect fully-automated the lumen contours in OCT images of a coronary artery pull-back procedure, it analyzes all available individual cross-sectional images, thus performs an analysis in three-dimensions (3D). The method also incorporates a “repair” algorithm that detects false detected contours, such as in case of side-branches or when a part of the artery is out of range, and fixes them. The automated method has been validated against human expert observers and showed no differences between them. However, the results of the automated method still need expert validation before creating a final report. There are still some possible corrections necessary, since 100% automated results is difficult to obtain as there are so many unexpected pathologies which could not all be foreseen. In this validation study it is reported that in 3% of the detected contours such manual interaction was necessary.
The advantage of OCT created and creates large enthusiasm and its proliferation into clinical practice is going rapidly. The developments of full-automated quantitative tools for QOCT could also make OCT an additional choice to be used in studies to improve the treatment of coronary atherosclerosis.
For more information see recent Article . Courtesy Nico Burining Thoraxcenter of the Erasmus MC, Rotterdam, The Netherlands.