HoliMAb: A holistic approach for Media–Adventitia border detection in intravascular ultrasound Francesco Ciompi a,b,⇑ , Oriol Pujol a,b , Carlo Gatta a,b , Marina Alberti a,b , Simone Balocco a,b , Xavier Carrillo c , Josepa Mauri-Ferre c , Petia Radeva a,b a Dept. of Applied Mathematics and Analysis, University of Barcelona, Spain b Computer Vision Center, Bellaterra, Spain c University Hospital Germans Trias i Pujol, Badalona, Spain article info Article history: Received 9 December 2011 Received in revised form 14 June 2012 Accepted 18 June 2012 Available online xxxx Keywords: Media–Adventitia border detection Intravascular ultrasound Multi-Scale Stacked Sequential Learning Error-correcting output codes Holistic segmentation abstract We present a fully automatic methodology for the detection of the Media–Adventitia border (MAb) in human coronary artery in Intravascular Ultrasound (IVUS) images. A robust border detection is achieved by means of a holistic interpretation of the detection problem where the target object, i.e. the media layer, is considered as part of the whole vessel in the image and all the relationships between tissues are learnt. A fairly general framework exploiting multi-class tissue characterization as well as contextual informa- tion on the morphology and the appearance of the tissues is presented. The methodology is (i) validated through an exhaustive comparison with both Inter-observer variability on two challenging databases and (ii) compared with state-of-the-art methods for the detection of the MAb in IVUS. The obtained averaged values for the mean radial distance and the percentage of area difference are 0.211 mm and 10.1%, respec- tively. The applicability of the proposed methodology to clinical practice is also discussed. Ó 2012 Elsevier B.V. All rights reserved. 1. Introduction Holism is a theory stating that parts of a whole are in intimate interconnection, such that they cannot exist independently of the whole, or cannot be understood without reference to the whole, which is thus regarded as greater than the sum of its part (Soanes and Angus, 2004). There exist, in computer vision, problems where the success in detection or segmentation of an object in an image highly depends on the context where the object is placed: the presence and the po- sition of specific regions around the object of interest give impor- tant insight on understanding the object itself. In medical imaging the difficulty in segmenting objects is even more remarkable, due to the weak definition of biological struc- tures and the presence of noise and artifacts depending on the image modality used for the diagnosis. In addition, in these cases the relationships and interconnections between regions are strictly ruled by the physical properties of tissues. The structure of the coronary artery is mainly defined by three regions: intima, media and adventitia (see Fig. 1a). The intima rep- resents the inner layer of the vessel, while the adventitia is the tis- sue constituting the outer part of the vessel, mainly composed of loose collagen and elastic tissue (Mintz, 2005). The media layer, characterized by a low collagen content, separates the intima from the adventitia. Two characteristics membranes can be defined, the Internal Elastic Lamina (IEL) and the External Elastic Lamina (EEL) (see Fig. 1b). The thickness of the media depends on the vessel con- dition. The typical thickness is of 200 lm in the absence of athero- sclerosis, and becomes thinner in the presence of disease (Mintz, 2005). In healthy arteries the shape of the vessel, delineated by the EEL, has a circular form, while in unhealthy cases it can become irregular depending on the plaque amount and positioning. The clinical condition of the artery is commonly characterized by parameters like the Plaque plus media Cross Sectional Area (CSA), the Maximum plaque plus media thickness and the Plaque Burden. All of these measurements require the delineation of the EEL or what is named as the Media–Adventitia border (MAb). Intravascular Ultrasound (IVUS) is a catheter-based imaging technique commonly used during percutaneous interventions. The IVUS image provides a high resolution tomographic recon- struction of the vessel appearance obtained by means of ultra- sound technology. The IVUS image allows to analyze the internal morphology of the artery (see Fig. 2a), making visible microscopic structures of tissue, like the media layer and the Media–Adventitia border. Due to its collagen content that makes it echo-lucent, the media layer in IVUS appears as a dark ring that surrounds the vessel; 1361-8415/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.media.2012.06.008 ⇑ Corresponding author at: Dept. of Applied Mathematics and Analysis, Univer- sitat de Barcelona, Gran Via de Les Corts Catalanes 585, 08007, Barcelona, Spain. E-mail address: fciompi@maia.ub.es (F. Ciompi). Medical Image Analysis xxx (2012) xxx–xxx Contents lists available at SciVerse ScienceDirect Medical Image Analysis journal homepage: www.elsevier.com/locate/media Please cite this article in press as: Ciompi, F., et al. HoliMAb: A holistic approach for Media–Adventitia border detection in intravascular ultrasound. Med. Image Anal. (2012), http://dx.doi.org/10.1016/j.media.2012.06.008