Michael Fieseler

Motion Correction of Whole-Body PET Data with a Joint PET-MRI Registration Functional

Respiratory motion is known to degrade image quality in PET imaging. The necessary acquisition time of several minutes per bed position will inevitably lead to a blurring effect due to organ motion. A lot of research has been done with regards to motion correction of PET data. As full-body PET-MRI became available recently, the anatomical data provided by MRI is a promising source of motion information. Current PET-MRI-based motion correction approaches, however, do not take into account the available information provided by PET data. PET data, though, may add valuable additional information to increase motion estimation robustness and precision.In this work we propose a registration functional that is capable of performing motion detection in gated data of two modalities simultaneously. Evaluation is performed using phantom data. We demonstrate that performing a joint registration of both modalities does improve registration accuracy and PET image quality.

Motion Correction in Respiratory Gated Cardiac PET/CT Using Multi-scale Optical Flow

Respiratory motion is a source of degradation in positron emission tomography. As the patients cannot hold breath during the PET acquisition, spatial blurring and motion artifacts are unavoidable which may lead to wrong quantification of the data. A solution based on respiratory-gating and optical flow based correction of the PET data is proposed. This includes deformation of the CT data for accurate attenuation and listmode based reconstruction. All methods are applied to real patient data and are evaluated with respect to three criteria.

Registration of depth and video data in depth image based rendering

Depth image based rendering (DIBR) has been proposed to create content for 3D-TV. In DIBR stereoscopic images are created from monoscopic images and associated depth data. Techniques deducing depth information from available video content have been applied to process video data lacking associated depth data for DIBR. Yet, artificial as well as recorded depth data may contain misalignments with respect to the video data. Misaligned depth data is a source of artifacts observable in the rendered 3D view. We show that by using an edge based registration method the spatial alignment of depth and video data can be improved, leading to an alleviation of the observed artifacts.

Motion correction in PET-MRI: A human torso phantom study

Respiratory and cardiac motion as degrading factors in PET imaging have been tackled by gating and subsequent PET-based motion detection techniques. The limitation of these approaches, however, is that only regions with high uptake yield sufficient motion information. Motion detection for tiny structures, e.g. arteriosclerotic plaques or small tumors, can prove difficult. In PET-MRI, the simultaneously acquired MR data are a promising source of motion information. In the present work, we give preliminary results for motion correction of PET data using information derived from MR data. A human torso phantom capable of simulating cardiac and respiratory motion was used to generate realistic data. Our preliminary results show that motion correction in PET-MRI is a promising approach.

Discontinuity-based registration of depth and video data in depth image based rendering

Depth image based rendering (DIBR) has been proposed to create content for 3D-TV. In DIBR, stereoscopic images are created from monoscopic images and associated depth data. Since for most of the available video content sensor depth data are lacking, methods to create artificial depth data for video content have been developed. Yet artificial as well as sensor depth data may contain misalignments with respect to video data. Misaligned depth data are a source of artifacts observable in rendered 3D views. We show that by using an edge-based registration method, the spatial alignment of depth and video data can be improved, leading to an alleviation of the observed artifacts.

Motion correction in Positron Emission Tomography considering Partial Volume Effects in optical flow estimation

Motion correction in Positron Emission Tomography (PET) using optical flow estimation can lead to image artifacts due to Partial Volume Effects (PVE). These artifacts appear especially in cardiac gated PET images and cause blurred edges in the averaged gates. In this paper we propose a new method to motion correct PET images considering the PVE during optical flow estimation. For this purpose we introduce a local intensity correction algorithm and combine it with the optical flow computation in an iterative scheme. The results of our approach show a qualitative and quantitative improvement of the motion corrected PET gates in examinations of both human patients and laboratory mice for pre-clinical research.

A multi-resolution optical flow based approach to respiratory motion correction in 3D PET/CT images

The problem of motion in PET/CT studies results from the fact that the PET images are formed over an elongated period of time whereas the CT acquisition takes only a few seconds. Additionally, the CT images are also used for attenuation correction of the PET data. The spatial discrepancy between the two sets of images, however, may lead to wrong attenuation and thus to wrong quantification of the radioactive uptake. We present a solution to this problem by respiratory-gating the PET data and correcting the PET images for motion with a multiresolution optical flow algorithm. Our algorithm is based on the combined local and global optical flow algorithm with modifications to allow for discontinuity preservation across organ boundaries and for application to 3D volume sets.

Fast 2D MRI acquisitions for motion correction in PET-MRI

We performed continuous, fast acquisitions of 2D MR slices covering the thorax under free breathing. In present work, acquired 2D stacks are re-stored using a respiration signal. The usage of 2D slices is similar to the method described in [1]. The proposed method, however, does not include a navigator and acquisition times are shorter. Dara were acquired from two patients on a Siemens Biograph mMR scanner (Siemens Healthcare, Erlangen, Germany) using a FLASH sequence, TE 1.38ms, TR 26ms, flip angle 12o, a 32-channel body-coil (acceleration factor of 8). 20 coronal slices were acquired with 3.9 x 6.25mm2 in-plane resolution (HF, LR), 128 x128 pixel, slice thickness 9mm, slice spacing 9mm, 26ms per 2D slice (total acquisition time 160s). A respiratory signal was estimated from affine registrations of an area showing respiratory motion and subsequently used to sort the acquired 2D stacks into 8 respiratory phases. Figure ​Figure11 shows input data and re-gated data. Noise is reduced by averaging of several 2D stacks. Additionally, cardiac motion is eliminated to a large extent, thus the generated dataset can be used for respiratory motion correction. Average correlation of 2D stacks assigned to a phase is 0.958 for dataset 1 (randomly selected gates: 0.933, SD 0.03). For dataset 2 the average correlation of 2D stacks assigned to a phase is 0.94 (randomly selected gates: 0.9, SD 0.036). Figure 1 Datasets a) & c) show 2D slice of two acquired datasets. b) & d) show a slice from the re-gated, averaged datasets. Note the reduction of noise in the regated dataset. The results indicate that the proposed method is suitable for use in respiratory motion correction of PET data. In future work we will evaluate our approach on more datasets. Additionally, we will use motion estimates for each acquired 2D image stack to correct motion frame-wise.

Motion estimation in PET-MRI based on dual registration: preliminary results for human data.

In current motion correction approaches in PET-MRI, motion information from PET data is neglected. We present an approach where PET and MRI data are used for motion estimation simultaneously. The presented approach has been evaluated on phantom data before [1]. Here, we present first results for human PET-MRI data. The registration functional for dual registration is given by 1 Here, RMR and RPET denote two reference volumes and TMR and TPET the template volumes to be registered, D is a distance functional, and S is a regularizer. The scalar value β allows to weight the influence of the data term for PET [1]. The functional has been implemented using the FAIR toolbox [3]. Five patients were scanned following a clinical FDG scan. A self-gated radial VIBE sequence [2] and PET Listmode data were acquired. The datasets were re-binned into 5 coinciding PET and MRI phases (gates). Registration were computed for β ∈ {0, 0.5, 1, 2}, α was chosen empirically as α = 20. Correlation coefficients were computed for the heart region. In Figure ​Figure2a2a we show correlation values for each gate of dataset 4. In all gates the correlation of the PET data is improved using the joint motion estimation approach using a weight of β = 2. In ​In2b2b average correlation values of all gates are shown for all datasets processed. Figure 1 Overlay of PET and MR data for dataset 4, first respiratory phase (gate). Figure 2 Correlation values for PET data. (a) Correlation values for the heart region in gates 2 to 5 for dataset number 4. (b) Average correlation values for all patients We have shown that using a joint motion estimation approach the correlation of PET data is improved compared to an estimation of the motion solely on MRI data. Currently, we are evaluating motion-correcting reconstructions using the motion estimates from the proposed method.

Motion correction of PET data with a joint registration functional for multi-modal data

Respiratory motion is a known image degradation factor in PET imaging. A lot of research has been done in the field of motion correction of PET data. With the recent development of full-body PET/MRI, the morphological data provided by MRI is a promising source of motion information. MRI-based motion correction of PET data is a topic of active research. These approaches, however, usually do not take into account the available information provided by PET data. In this work we show results of a phantom study using a joint motion estimation approach that incorporates motion information from both PET and MRI data.