Alvén, J; Heurling, K; Smith, R; Strandberg, O; Schöll, M; Hansson, O; Kahl, F; (2019) A Deep Learning Approach to MR-less Spatial Normalization for Tau PET Images. In: Shen, D and Liu, T and Peters, TM and Staib, LH and Essert, C and Zhou, S and Yap, PT and Khan, A, (eds.) Medical Image Computing and Computer Assisted Intervention – MICCAI 2019: 22nd International Conference, Shenzhen, China, October 13–17, 2019, Proceedings, Part II. (pp. pp. 355-363). Springer: Cham, Switzerland.

Preview

Text Alvén_MICCAI2019-camerareadypaper.pdf - Published Version Download (363kB) | Preview

Abstract

The procedure of aligning a positron emission tomography (PET) image with a common coordinate system, spatial normalization, typically demands a corresponding structural magnetic resonance (MR) image. However, MR imaging is not always available or feasible for the subject, which calls for enabling spatial normalization without MR, MR-less spatial normalization. In this work, we propose a template-free approach to MR-less spatial normalization for [18F]flortaucipir tau PET images. We use a deep neural network that estimates an aligning transformation from the PET input image, and outputs the spatially normalized image as well as the parameterized transformation. In order to do so, the proposed network iteratively estimates a set of rigid and affine transformations by means of convolutional neural network regressors as well as spatial transformer layers. The network is trained and validated on 199 tau PET volumes with corresponding ground truth transformations, and tested on two different datasets. The proposed method shows competitive performance in terms of registration accuracy as well as speed, and compares favourably to previously published results.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item