doi:10.1016/j.ultrasmedbio.2007.02.013 ● Original Contribution PIXEL COMPOUNDING: RESOLUTION-ENHANCED ULTRASOUND IMAGING FOR QUANTITATIVE ANALYSIS Z HI Y ANG , § T HERESA A. T UTHILL , † D AVID L. R AUNIG , ‡ M ARTIN D. F OX ,* AND M OSTAFAANALOUI ‡ *University of Connecticut, Storrs, CT, USA; † VirtualScopics, Rochester, NY, USA; ‡ Pfizer Global Research and Development, Groton, CT, USA; and § University of Michigan, Ann Arbor, MI,USA Abstract—Accurate measurement of structural features represented in medical images is important trials and patient diagnosis. A key factor for precision is spatial resolution, which in ultrasonic imag by transducer array arrangements, transmitting frequency, and data acquisition firmware. In this paper, a variation of pixel compounding is proposed to enhance ultrasound resolution using acquired cine loops. The technique operates on a sequence of ultrasound B-scan images acquired with random motion. Subpixel registration is estimated and a maximum a posteriori (MAP) approach with the shift information is u reconstruct a high-resolution single image. A nonhomogeneous anisotropic diffusion algorithm follo the estimation process and is implemented to enhance the high-resolution edges. Preliminary tests using simulations and phantom studies show promising results. Pixelcompounding can be a powerful prepro- cessing tool to assure accurate segmentation, measurement, and analysis of ultrasound images. (E-mail: theresa.tuthill@pfizer.com) © 2007 World Federation for Ultrasound in Medicine & Biology. Key Words: Ultrasound image, Pixel compounding, Anisotropic diffusion, Super-resolution reconstruction. INTRODUCTION Ultrasound imaging has been and continues to be an important medical imaging modality due to its low cost and high safety profile. However, low spatial resolution and speckle noise limit the use of quantitative ultrasound imaging. Researchers have proposed spatial compound- ing (Jespersen et al.1998;Shankar 1996; Trahey et al. 1986),frequencycompounding (Amir et al. 1986; Magnin et al.1982;Saniie et al.1992;Silverstein and O’Donnell 1985;Trahey etal. 1986) and a variety of adaptive filtering techniques (Abd-Elmoniem et al. 2002; Czerwinski et al. 1999; Hao et al. 1999; Karaman et al. 1995; Lee 1980; Loupas et al. 1989; Yang and Fox 2004; Yang etal.2003;Yu and Acton 2002) to improve the signal-to-noise of the image. With spatial compounding, often only the lateral resolution is improved while the depth resolution is compromised. In frequency com- pounding, the resolution improvement is constrained by the highest frequency used. Higher ultrasound frequen- cies are limited in their utility due to the attenuation of high frequency ultrasound in tissue. Adaptive filtering techniques are much more effective in terms of noise re duction and discontinuity-preservation, but the resolutio processed images is always lower than the originals due the low-pass nature of the noise reduction filters. Image restoration can remove the blurring effect in images, but requires the knowledge of the point spread function, an resolution will not reach the subpixel accuracy. In this paper, a technique is proposed to achieve a subpixel-accuracy reconstruction of ultrasound images. Pixelcompounding parallels the more common spatial compounding and frequency compounding and uses the additional intensity information available from random movements of edges within the image. Images in the sequence should have slight, subpixel shifts due to eithe transducer orsubjectmotion.Since the intensity at a particular pixel is given by the integral of the incident signaloverthe area ofthe pixelpatch,the intensity weight center of a pixel for a feature that is smaller tha a pixelshould be a subpixel location within the pixel. The goalof pixelcompounding is to use changes in intensity due to subpixel shifts to estimate the subpixel location at a finer imaging grid so that a super-resolved and detail-revealing image can be reconstructed. The pixelcompounding described here is imple- mented from a maximum a posteriori (MAP) formulation Address correspondence to: Theresa Tuthill, VirtualScopics/Pfizer Global Research and Development, MS 8260-2301, Eastern Point Road, Groton CT 06340-2301. E-mail: theresa.tuthill@pfizer.com Ultrasound in Med. & Biol.,Vol. 33,No.8, pp.1309 –1319, 2007 Copyright © 2007 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/07/$–see front matter 1309