78P - Registration-based automated lesion detection and therapy evaluation of tumors in whole body PET-MR images

Date 11 September 2017
Event ESMO 2017 Congress
Session Poster display session
Topics Imaging, Diagnosis and Staging
Translational Research
Presenter Håkan Ahlström
Citation Annals of Oncology (2017) 28 (suppl_5): v1-v21. 10.1093/annonc/mdx361
Authors H. Ahlström1, S. Ekström1, T. Sjöholm1, R. Strand1, J. Kullberg1, E. Johansson2, P. Hagmar2, F. Malmberg1
  • 1Division Of Radiology, Department Of Surgical Sciences, Uppsala University, 751 85 - Uppsala/SE
  • 2-, Antaros medical, 431 83 - Mölndal/SE

Abstract

Background

Integrated PET/MR scanners can simultaneously acquire whole body functional PET data together with morphological and functional MR data. Whole-body PET-MRI datasets contain huge amounts of spatially detailed morphological, functional and metabolic information. We propose a method, based on deformable registration to a whole-body atlas, for computer aided detection of lesions in image data from an integrated PET-MRI system.

Methods

Images were acquired using an integrated 3T PET-MRI system (Signa PET/MR, GE Healthcare). Fat and water MR images were collected using a Dixon MR Attenuation Correction (MRAC) sequence (TE 1.67ms, TR 4.05ms, voxel size: 2x2x5.2 mm). Subjects underwent a PET scan after injection of [F18]-FDG (2 MBq/kg) with 3 minutes per bed, with a 100-120 minute interval between injection and scan start. PET reconstruction was performed using GE’s fully 3D Time-of-Flight iterative reconstruction (2 iterations, 28 subsets, standard 5 mm filter, voxel size 3.125x3.125x2.78). Deformable image registration was used to spatially align subjects to a previously created morpphological and functional whole-body imaging atlas (Ekström et al., ISMRM 17), to allow voxel-wise comparisons between the imaged subjects and the atlas. Each voxel in the atlas contains mean and standard deviation of the PET uptake. Utilizing the knowledge that low ADC-values (low diffusion measured by MRI) and high FDG uptake (high metabolism measured by PET) is indicative of malignancy, suspected lesions can be detected by measuring how much the FDG uptake in each voxel deviates from “normality”, as defined by the atlas. This approach generates a voxel-wise “lesion probability map” for the imaged subject. The same registration approach can be used to quantify longitudinal changes in detected lesions, for treatment evaluation.

Results

Lesion probability maps have been generated for patients with manually identified lesions, correctly assigning high values to the regions manually identified as suspected lesions.

Conclusions

The proposed method is promising for lesion detection in whole body PET-MRI images. Future work includes quantitative verification of the accuracy of the detected regions, comparing against manual detection by a radiologist.

Clinical trial identification

Legal entity responsible for the study

Uppsala University

Funding

Antaros Medical

Disclosure

H. Ahlström, J. Kullberg: Co-founder and owner of Antaros Medical.

rn

All other authors have declared no conflicts of interest.

Disclosure

H. Ahlström, J. Kullberg: Co-founder and owner of Antaros Medical.

\r\n

All other authors have declared no conflicts of interest.

Disclosure

H. Ahlström, J. Kullberg: Co-founder and owner of Antaros Medical.

All other authors have declared no conflicts of interest.