Application of independent component analysis to dynamic contrast-enhanced imaging for assessment of cerebral blood perfusion X.Y. Wu a,b, * , G.R. Liu a,c a Centre for Advanced Computations in Engineering Science, National University of Singapore, BLK EA-04-26, 10 Kent Ridge Crescent, Singapore 119260, Singapore b Institute of Engineering Science, National University of Singapore, University Hall, #UHL-05, Lee Kong Chian wing, 21 Lower Kent Ridge Road, Singapore 119077, Singapore c Mechanical Engineering Department, National University of Singapore, 9 Engineering Drive 1, Singapore 117576, Singapore Received 8 November 2005; received in revised form 23 January 2007; accepted 21 March 2007 Available online 30 March 2007 Abstract Dynamic contrast-enhanced (DCE) imaging is widely used for in vivo assessment of the cerebral blood perfusion. In this work, we investigate the use of independent component analysis (ICA) on DCE imaging data for assessment of cerebral blood perfusion, without any prior knowledge of the underlying tissue vasculature and arterial input function. The minimum description length (MDL) criterion and principle component analysis (PCA) were employed to reduce the dimension of the data. An oscillating index method was used to select the components of interest. Numerical simulation and patient case studies were carried out to investigate the performance of ICA. The results show that ICA is able to extract physiologically meaningful components from the DCE imaging data. The advantages of ICA include its efficiency of computation, clarity of obtained component maps, and no need of the manually selected input function. The obtained independent component maps can provide reliable reference to identify the arterial and venous structure, and allow better demarcation of the tumor territories. The potential of ICA to be a useful clinical tool for diagnosis of cerebral vascular disease and for the assessment of treatment response has been demonstrated. Ó 2007 Elsevier B.V. All rights reserved. Keywords: Independent component analysis; Dynamic contrast-enhanced imaging; Cerebral blood perfusion 1. Introduction Dynamic contrast-enhanced (DCE) imaging with com- puted tomography (CT) or magnetic resonance imaging (MRI) is widely used in scientific research and clinical prac- tice for in vivo assessment of the cerebral blood perfusion (Klotz and Konig, 1999; Koenig et al., 2001; Vonken et al., 1999; Wirestam et al., 2000). Increasing evidence has shown that the perfusion information derived from DCE imaging data can potentially be helpful to understand the pathophysiology of cerebral vascular diseases (Cala- mante et al., 2002; Koenig et al., 2001), and allows the cli- nicians to monitor and assess the therapeutic effects (Guckel et al., 1994; Koh et al., 2004). In the technique of DCE imaging, an injection of con- trast medium (tracer) is administrated intravenously, and then continuous acquisition of data from a single scan slice is performed. The passage of tracer particles through the brain creates the temporal changes of physical signals (amount of absorbed X-ray radiation in CT; local suscep- tibility in MRI), which in turn cause intensity changes on the captured sequential images. From those images, we can assess the changes in tracer concentration over time, which not only monitor the tracer kinetic behavior but also reflect the brain hemodynamic status. In clinical practice, DCE imaging normally acquires 50–100 images per case with a scan time of 1 s/frame or 0.5 s/frame. To extract 1361-8415/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.media.2007.03.005 * Corresponding author. Address: Centre for Advanced Computations in Engineering Science, National University of Singapore, BLK EA-04-26, 10 Kent Ridge Crescent, Singapore 119260, Singapore. E-mail address: xingyewu@pmail.ntu.edu.sg (X.Y. Wu). www.elsevier.com/locate/media Medical Image Analysis 11 (2007) 254–265