TY - GEN
T1 - Longitudinal guided super-resolution reconstruction of neonatal brain MR images
AU - Shi, Feng
AU - Cheng, Jian
AU - Wang, Li
AU - Yap, Pew Thian
AU - Shen, Dinggang
N1 - Publisher Copyright:
© Springer International Publishing Switzerland 2015.
PY - 2015
Y1 - 2015
N2 - Neonatal images have low spatial resolution and insufficient tissue contrast. Generally, interpolation methods are used to upsample neonatal images to a higher resolution for more effective image analysis. However, the resulting images are often blurry and are susceptible to partial volume effect. In this paper, we propose an algorithm that utilizes longitudinal prior information for effective super-resolution reconstruction of neonatal images. We use a non-local approach to learn the spatial relationships of brain structures in highresolution longitudinal images and apply this information to the superresolution reconstruction of the neonatal image. In other words, the recurring patterns throughout the longitudinal scans are leveraged for reconstructing the neonatal image with high resolution. To solve this otherwise ill-posed inverse problem, low-rank and total-variation regularizations are enforced. Experiments performed on both T1- and T2-weighted MR images of 28 neonates demonstrate that the proposed method is capable of recovering more structural details and outperforms methods such as nearest neighbor interpolation, spline-based interpolation, non-local means upsampling, and both low-rank and total variation based super-resolution.
AB - Neonatal images have low spatial resolution and insufficient tissue contrast. Generally, interpolation methods are used to upsample neonatal images to a higher resolution for more effective image analysis. However, the resulting images are often blurry and are susceptible to partial volume effect. In this paper, we propose an algorithm that utilizes longitudinal prior information for effective super-resolution reconstruction of neonatal images. We use a non-local approach to learn the spatial relationships of brain structures in highresolution longitudinal images and apply this information to the superresolution reconstruction of the neonatal image. In other words, the recurring patterns throughout the longitudinal scans are leveraged for reconstructing the neonatal image with high resolution. To solve this otherwise ill-posed inverse problem, low-rank and total-variation regularizations are enforced. Experiments performed on both T1- and T2-weighted MR images of 28 neonates demonstrate that the proposed method is capable of recovering more structural details and outperforms methods such as nearest neighbor interpolation, spline-based interpolation, non-local means upsampling, and both low-rank and total variation based super-resolution.
UR - http://www.scopus.com/inward/record.url?scp=84927943734&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-14905-9_6
DO - 10.1007/978-3-319-14905-9_6
M3 - Conference contribution
AN - SCOPUS:84927943734
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 67
EP - 76
BT - Spatio-temporal Image Analysis for Longitudinal and Time-Series Image Data - 3rd International Workshop, STIA 2014 Held in Conjunction with MICCAI 2014, Revised Selected Papers
A2 - Durrleman, Stanley
A2 - Fletcher, Tom
A2 - Gerig, Guido
A2 - Niethammer, Marc
A2 - Pennec, Xavier
PB - Springer Verlag
T2 - 3rd International Workshop on Spatio-temporal Image Analysis for Longitudinal and Time-Series Image Data, STIA 2014 in conjunction with the International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2014
Y2 - 18 September 2014 through 18 September 2014
ER -