Organ-Specific Context-Sensitive Single and Dual Energy CT (DECT) Image Reconstruction, Display and Analysis

Size: px

Start display at page:

Download "Organ-Specific Context-Sensitive Single and Dual Energy CT (DECT) Image Reconstruction, Display and Analysis"

Ethel James
5 years ago
Views:

Organ-Specific Context-Sensitive Single and Dual Energy CT (DECT) Image Reconstruction, Display and Analysis Sabrina Dorn 1, Shuqing Chen², Francesco Pisana 1, Joscha Maier 1, Michael Knaup 1, Stefan

1 Organ-Specific Context-Sensitive Single and Dual Energy CT (DECT) Image Reconstruction, Display and Analysis Sabrina Dorn 1, Shuqing Chen², Francesco Pisana 1, Joscha Maier 1, Michael Knaup 1, Stefan Sawall 1, Andreas Maier², Michael Lell³, and Marc Kachelrieß 1 1 German Cancer Research Center (DKFZ), Heidelberg, Germany ²Friedrich-Alexander University Erlangen-Nürnberg, Germany ³Hospital Nürnberg, Paracelsus Medical University

display and analysis using prior anatomical

2 Aim To combine mutually exclusive CT image properties into a single organ-specific image reconstruction, display and analysis using prior anatomical information. smooth kernel reconstruction sharp kernel reconstruction 2

3 Aim To combine mutually exclusive CT image properties into a single organ-specific image reconstruction, display and analysis using prior anatomical information. sliding thin slab (STS) display lung window STS MIP body window 0.5 mm slab 10 mm slab 3

4 Aim To combine mutually exclusive CT image properties into a single organ-specific image reconstruction, display and analysis using prior anatomical information. Rho/Z Optimum Contrast Calculi Characterization Bone Marrow Xenon Heart PBV Direct Angio Lung Analysis Monoenergetic Plus Lung Nodules Virtual Unenhanced Hardplaque Display Syngo.CT DECT application examples. Virtual unenhanced contains liver VNC, lung analysis contains lung PBV. Courtesy of Siemens Healthineers, Forchheim, Germany 4

Method Prior anatomical knowledge: 3D fully convolutional network 1 Segmentation of dual energy data Cascaded neural network architecture 1. Detection of the region of interest (ROI) 2.

5 Method Prior anatomical knowledge: 3D fully convolutional network 1 Segmentation of dual energy data Cascaded neural network architecture 1. Detection of the region of interest (ROI) 2. Final detection of organ boundaries I L, I prediction H detection of ROIs detection of multiple organs Automatic segmentation: liver, kidneys, spleen, lung, bone, aorta. Thresholding remaining voxels into the following tissue types: muscles, fat, vasculature. Currently, manual corrections are necessary (until today). [1] S. Chen, H. Roth, S. Dorn, M. May, A. Cavallaro, M. Lell, M. Kachelrieß, H. Oda, K. Mori, and A. Maier. Towards Automatic Abdominal Multi-Organ Segmentation in Dual Energy CT using Cascaded 3D Fully Convolutional Network. CoRR,

6 Method Segmentation delivers a binary mask for each organ. 1. Smoothing of the binary masks to cope with the boundaries of adjacent anatomical structures. 2. Use smoothed masks to allow for individual settings for each organ. Context-sensitive (CS) resolution CS display CS dual energy evaluation Context-sensitive (CS) = organ-dependent parameter adaptation 6

7 Context-Sensitive Resolution standard low resolution image (smooth kernel) resolution-mixed image (high resolution in lung and bone, low noise in soft tissue) CS resolution standard high resolution image (sharp kernel) 7

tissue) CS resolution standard high resolution image (sharp kernel)

8 Context-Sensitive Resolution standard low resolution image (smooth kernel) resolution-mixed image (high resolution in lung and bone, low noise in soft tissue) CS resolution standard high resolution image (sharp kernel) increased spatial resolution in bone and lung decreased noise level in soft tissue 8

9 Context-Sensitive Display conventional windowing CS resolution bone window body window lung window 9

10 Context-Sensitive Display CS resolution CS display windowing 10

11 Context-Sensitive Display CS resolution CS display windowing 11

12 Context-Sensitive Display CS resolution CS display windowing 12

13 Context-Sensitive Display CS resolution CS display windowing 13

14 Context-Sensitive Display CS resolution CS display windowing 14

15 Context-Sensitive Display CS resolution CS display windowing STS mean in soft tissue (5 mm) 15

16 Context-Sensitive Display CS resolution CS display windowing STS mean in soft tissue (5 mm) STS MIP in lung (10 mm) 16

17 Bone marrow Iodine overlay Lung PBV Context-Sensitive Dual Energy 100 HU 10 6 mg/ml mg/ml simultaneous DE evaluation with commonly used applications Calcium-oxalate-stone -100 HU 0 mg/ml 0 mg/ml Uric acid-stone 17

18 Bone marrow Iodine overlay Lung PBV Context-Sensitive Dual Energy ROI 1 (lung): 100 HU 10 6 mg/ml mg/ml ROI 1 11 % CM: 0.7 mg/ml 1 % 88 % ROI 2 ROI 2 (liver): 29 % CM: 2.9 mg/ml 8 % -100 HU 0 mg/ml 0 mg/ml 63 % 3 material decomposition with organ-dependent basis materials : volume fractions 18

Conclusion Method strongly depends on segmentation accuracy still needs improvement Context-sensitive resolution-mixing combines mutually exclusive image properties» high spatial resolution in bone

19 Conclusion Method strongly depends on segmentation accuracy still needs improvement Context-sensitive resolution-mixing combines mutually exclusive image properties» high spatial resolution in bone and lung» low noise in soft tissue Context-sensitive display able to present significantly more information to the reader simultaneously Organ-specific DE evaluation potential to facilitate the diagnosis 19

20 Thank You! This study was supported by the Deutsche Forschungsgemeinschaft (DFG) under grant KA 1678/20-1, LE 2763/2-1 and MA 4898/5-1. This presentation will soon be available at Job opportunities through DKFZ s international PhD or Postdoctoral Fellowship programs ( or directly through Marc Kachelriess (marc.kachelriess@dkfz.de). Parts of the reconstruction software were provided by RayConStruct GmbH, Nürnberg, Germany. 20

Preclinical X-Ray Computed Tomography

Preclinical X-Ray Computed Tomography Marc Kachelrieß German Cancer Research Center (DKFZ) Heidelberg, Germany www.dkfz.de/ct Is CT a Molecular Imaging Modality? Imaging Modality Molecular Sensitivity