mcet talk topomorpho

Post on 16-Apr-2017

224 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Subhadip Basu, Ph.D.

Department of Computer Sc. & Engineering,

Jadavpur University, Kolkata, INDIA

A theory and algorithm for separating two structures sharing a common intensity band and conjoined at different unknown locations and scales, is presented

The method is applied for segmenting vasculature in patients with intracranial aneurysms via CT angiography (CTA)

05-04-2014 2 S. Basu,Jadavpur University, INDIA

The segmentation for bone and vessels combines fuzzy distance transform and fuzzy connectivity to iteratively open two objects starting at large scales and progressing toward smaller scales

The accuracy of the method has been examined both qualitatively and quantitatively on mathematically generated phantoms, CT images of a pig pulmonary vessel cast

phantom, and cerebral CT angiography images of human

subjects

05-04-2014 3 S. Basu,Jadavpur University, INDIA

An axial image slice from a

human CT angiogram

05-04-2014 4 S. Basu,Jadavpur University, INDIA

Intensity-based membership functions for vessel (red) and bone (green) along with pure and shared intensity bands.

05-04-2014 5 S. Basu,Jadavpur University, INDIA

Multi-scale fusion of bone and vessel demands a locally adaptive multi-scale opening

05-04-2014 6 S. Basu,Jadavpur University, INDIA

During an iteration, it opens two structures over a specific scale range by:

extending object separation from previous iteration using the optimum opening structure defined by FDT and fuzzy connectivity, and

dilate the two separated objects using constrained dilation

05-04-2014 7 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 8

05-04-2014 S. Basu,Jadavpur University, INDIA 9

05-04-2014 S. Basu,Jadavpur University, INDIA 10

05-04-2014 S. Basu,Jadavpur University, INDIA 11

After Local Normalization, FDT values lie within the interval [0,1].

Local scale is defined as the depth (i.e., the FDT value) at the nearest

locally-deepest voxels.

05-04-2014 S. Basu,Jadavpur University, INDIA 12

05-04-2014 S. Basu,Jadavpur University, INDIA 13

A

Lower FDT value Higher FDT value

Strongest path between A

and B. FDT value of the

weakest point is higher than

the other path

Not the strongest path

between the A and B

SA SB

05-04-2014 S. Basu,Jadavpur University, INDIA 14

SA SB

Undecided region, having FC Strength, γA= γB Strongest path between SA,SB

RB : Region

assigned to SB RA : Region

assigned to SA

15 05-04-2014 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 16

05-04-2014 S. Basu,Jadavpur University, INDIA 17

Morphologically dilated RA Morphologically dilated RB

Morphological neighborhood

Both separated regions

radially expand over

morphological neighborhood

until stopped by each other

Cross-sectional views

before after

reconstruction 18 05-04-2014 S. Basu,Jadavpur University, INDIA

After morphological reconstruction, the hollow annular region is filled in

Now, we are ready to expand the separation to the next finer scale

We start with the result of previous separation use it to determine seeds of individual objects

05-04-2014 19 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 20

05-04-2014 S. Basu,Jadavpur University, INDIA 21

05-04-2014 S. Basu,Jadavpur University, INDIA 22

05-04-2014 S. Basu,Jadavpur University, INDIA 23

To generate a vessel cast data, the animal was first exsanguinated.

While maintaining ventilation at low PEEP, the pulmonary vasculature was flushed with 1L 2% Dextran solution and pneumonectomy was performed.

While keeping the lungs inflated at approximately 22 cm H2O Pawy, a rapid hardening methyl methacrylate compound (Orthodontic Resin, DENTSPLY International, York, PA) was injected into the vasculature to create a cast of the pulmonary arterial and venous trees.

The casting compound was mixed with red oil paint for the venous (oxygenated) side and blue oil paint for the arterial (deoxygenated) side of the vascular beds.

Data courtesy: Dept. of Radiology, UIowa

05-04-2014 S. Basu,Jadavpur University, INDIA 24 (a)

05-04-2014 S. Basu,Jadavpur University, INDIA 25

Axial and coronal image slices from the original CT image of the phantom with different

contrast for A/V trees. CT intensity-based classification of artery and vein where the effect of

partial voluming appears as thin red films wrapping around blue arteries.

05-04-2014 S. Basu,Jadavpur University, INDIA 26

CT intensity histogram of the phantom, where the two CT intensity values Imin and

Iartery segments the background and pure artery regions

05-04-2014 S. Basu,Jadavpur University, INDIA 27

Optimum thresholding

MSO algorithm

05-04-2014 S. Basu,Jadavpur University, INDIA 28

(b) (c) Optimum thresholding MSO algorithm

CT angiogram data sets were collected using Siemens Somatom Sensation 16 scanner at 120 KV, rotation time of 0.5 sec, 0.75 pitch and 0.75 mm collimation. The contrast medium used was 75 cc of Omipaque 300.

Data courtesy: Dept. of BME,

Univ. of Iowa.

29 05-04-2014 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 30

05-04-2014 S. Basu,Jadavpur University, INDIA 31

CTA intensity histogram values Imin and Ibone segmenting the background and the pure

bone regions

32

Cerebral

CTA Slice

05-04-2014 S. Basu,Jadavpur University, INDIA

33

Fused bone and

vessel shown in

RED

Pure bone shown

in GREEN

05-04-2014 S. Basu,Jadavpur University, INDIA

34

Bone-Vessel

segmentation after

1st iteration

05-04-2014 S. Basu,Jadavpur University, INDIA

35

Bone-Vessel

segmentation after

2nd iteration

05-04-2014 S. Basu,Jadavpur University, INDIA

36

Bone-Vessel

segmentation after

final iteration

05-04-2014 S. Basu,Jadavpur University, INDIA

37

Cerebral

vasculature

segmented from

bone

Anearysm

05-04-2014 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 38

(b) (a) (c)

05-04-2014 S. Basu,Jadavpur University, INDIA 39

Large scale objects

05-04-2014 S. Basu,Jadavpur University, INDIA 40

Small scale objects

41

Data#1 Data#2 Data#3 Data#4 Data#5

Error Large Vessel 1.688103 0.767222 2.925486 0.689809 0.209512

Error Small Vessel 1.339764 0.21221 0.361384 0.46729 2.053216

Error Large Bone 0.45552 0.375449 1.359491 0.267023 2.325581

Error Small Bone 0.133976 0.163239 0.103252 0.489542 0.398072

Total Error 3.617363 1.518119 4.749613 1.913663 4.986382

05-04-2014 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 42

To estimate intra-user reproducibility, we define “agreement” as:

AGR = (V1V2) (B1B2)/(V1V2 B1 B2) where V1 and V2 are volumes rendered from the set of

seeds marked by the user on two different runs of experiment

43

Average agreement is 94.2 ± 3.8%.

05-04-2014 S. Basu,Jadavpur University, INDIA

05-04-2014 S. Basu,Jadavpur University, INDIA 44

(a) (b)

(c) (d)

Scanner: Siemens Sensation 64

MDCT scanner

CT Parameters: 120 kVp and

100 mAs.

Scanned at 0.75 mm slice

thickness

Reconstructed at 0.5 mm slice-

thickness and 0.6x0.6mm2 in-

plane resolution.

Artery and Vein structures are

inseparable in intensity space

45

Image courtesy: Prof. Punam K. Saha, Univ. of Iowa, USA

46

Mutually blinded inter-

user reproducibility

Agreement: 93%

Image courtesy: Prof. Punam K. Saha, Univ. of Iowa, USA

47

Mutually blinded inter-

user reproducibility

Agreement: 91%

Image courtesy: Prof. Punam K. Saha, Univ. of Iowa, USA

A novel approach for multi-scale opening in shared intensity space

A/V tree and B/V separation problems are solved using the developed

MSO algorithm

Results on computer-generated phantoms show high accuracy

Promising results on pig lung phantom and human cerebral CTA data

Possible extensions to problems with multi scale separation of objects

48 05-04-2014 S. Basu,Jadavpur University, INDIA

Prof. Punam K Saha, Dept. of ECE, Univ. of Iowa

Prof. Eric Hoffman, Dept. of Radiology, Univ. of Iowa

Prof. M. L. Raghavan, Dept. of BME, Univ. of Iowa

Dr. Robert E. Harbaugh, Penn State Hershey Medical Center

My visit to the Structural Imaging Laboratory, Univ. of Iowa, USA, was funded by the BOYSCAST fellowship (SR/BY/E-15/09), Dept. of Science and Technology, Govt. of INDIA.

This study is supported in part by the FASTTRACK grant (SR/FTP/ETA-04/2012) by DST, Govt. of India.

49 05-04-2014 S. Basu,Jadavpur University, INDIA

P.K. Saha, Z. Gao, S.K. Alford, M. Sonka, and E.A. Hoffman, “Topomorphologic separation of fused isointensity objects via multiscale opening: separating arteries and veins in 3-D pulmonary CT.,” IEEE Transactions on Medical Imaging, vol. 29, 2010, pp. 840-851

S. Basu, M. L. Raghavan, E. A. Hoffman, P. K. Saha, “Multi-scale opening of conjoined structures with shared intensities: methods and applications,” in Proc. IEEE International conference on Intelligent Computation and Bio-Medical Instrumentation (ICBMI 2011), Wuhan, China, December 14 - 17, , pp. 128-131, 2011.

S. Basu, M. L. Raghavan, P. K. Saha, “Vascular segmentation in CT angiography for patients with intracranial aneurysms using a new multi-scale opening algorithm,” in Proc. of International conference on Bio-Medical Engineering (ICBME 2011), Manipal, India, pp. 252-257, December 10 - 12, 2011.

Z. Gao, R. W. Grout, C. Holtze, E. A. Hoffman, and P. Saha, “A New Paradigm of Interactive Artery/Vein Separation in Noncontrast Pulmonary CT Imaging Using Multiscale Topomorphologic Opening.,” IEEE transactions on bio-medical engineering, vol. 59, no. 11, pp. 3016–27, Nov. 2012.

50 05-04-2014 S. Basu,Jadavpur University, INDIA

top related