shape modeling of the paranasal sinuses to estimate natural - - PowerPoint PPT Presentation

β–Ά
shape modeling of the paranasal
SMART_READER_LITE
LIVE PREVIEW

shape modeling of the paranasal sinuses to estimate natural - - PowerPoint PPT Presentation

Jan 29, 2024 200 likes β€’505 views

Automatic segmentation and statistical shape modeling of the paranasal sinuses to estimate natural variations Ayushi Sinha a , Simon Leonard a , Austin Reiter a , Masaru Ishii b , Russell H. Taylor a and Gregory D. Hager a a Dept. of Computer

slide-1
SLIDE 1

Automatic segmentation and statistical shape modeling of the paranasal sinuses to estimate natural variations

Ayushi Sinhaa, Simon Leonarda, Austin Reitera, Masaru Ishiib, Russell H. Taylora and Gregory D. Hagera

  • aDept. of Computer Science, The Johns Hopkins University, Baltimore, MD 21218, USA
  • bDept. of Otolaryngology-Head and Neck Surgery, Johns Hopkins Medical Institutions, Baltimore, MD 21287, USA
slide-2
SLIDE 2

Introduction

  • Functional endoscopic sinus surgery (FESS) is a

routine operation performed by an otolaryngologist

  • Between 200,000 and 600,000 endoscopic interventions

per year in the USA[1][2][3]

  • Navigation during surgery can be improved using

pre-operative CT

  • Reduces likelihood of potential complications
  • Enhances patient safety and outcome

[1]Hosemann W, Draf C. Danger points, complications and medico-legal aspects in endoscopic sinus

  • surgery. GMS Current Topics in Otorhinolaryngology, Head and Neck Surgery. 2013;12:Doc06.

[2]Hepworth EJ, Bucknor M, Patel A, Vaughan WC. Nationwide survey on the use of image-guided functional

endoscopic sinus surgery. Otolaryngol Head Neck Surg. 2006 Jul;135(1):68–73.

[3]Psaltis AJ, Soler ZM, Nguyen SA, Schlosser RJ. Changing trends in sinus and septal surgery, 2007 to 2009. Int

Forum Allergy Rhinol. 2012 Sep-Oct;2(5):357–361.

slide-3
SLIDE 3

Nasal Cycle

  • Alternating partial congestion and

decongestion of the nasal cavities due to the expansion and contraction of the inferior, middle, and superior turbinates[4]

  • Each cycle can span between ~50 minutes

to several hours[5]

  • Need to compensate for this regularly

deforming topology

[4]Hasegawa M, Kern EB, The human nasal cycle. Mayo Clinic Proceedings. May 1977;51:28-34 [5]Atanasov AT. Length of Periods in the Nasal Cycle during 24-Hours Registration. Open Journal of Biophysics.

2014;4:93-96

slide-4
SLIDE 4

Can we estimate this deformation?

  • Lack of longitudinal studies
  • But, plenty of head CTs from different

patients

  • Can we characterize this deformation from

head CTs of a large population?

slide-5
SLIDE 5

Can we estimate this deformation?

  • Hypothesis

is: Given CTs of π‘œ individuals, it is likely that the turbinates of each individual are at a different state in the nasal cycle than all others. Therefore, a statistical model of the turbinates built from these π‘œ CTs should also reflect natural variation.

slide-6
SLIDE 6

Statistical Shape Model (SSM) Patient CTs Template

Method

Deformed Template

Deformably Register Deform Template PCA

slide-7
SLIDE 7

Template Creation

[6]

[6]BB Avants, P Yushkevich, J Pluta, D Minko, M Korczykowski, J Detre, JC Gee, β€œThe optimal template effect in

hippocampus studies of diseased populations," NeuroImage 49(3), p. 2457, 2010.

  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

slide-8
SLIDE 8

Automatic Segmentation

[7]

Template Mesh

Deformation Fields

Deformed Meshes

[7]BB Avants, NJ Tustison, . Song, PA Cook, A Klein, and JC Gee, β€œA reproducible evaluation of ANTs similarity

metric performance in brain image registration," NeuroImage 54(3), pp. 2033-2044, 2011.

  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

slide-9
SLIDE 9

Deformable Registration (DR)

[7]

[7]BB Avants, NJ Tustison, . Song, PA Cook, A Klein, and JC Gee, β€œA reproducible evaluation of ANTs similarity

metric performance in brain image registration," NeuroImage 54(3), pp. 2033-2044, 2011.

  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

slide-10
SLIDE 10

Gradient Vector Flow (GVF)

[10][11]

[10] C. Xu and J. L. Prince, β€œGradient vector ow: A new external force for snakes," in Computer Vision and

Pattern Recognition, IEEE Computer Society Conference on, pp. 66-71, 1997.

[11] C. Xu and J. Prince, β€œSnakes, shapes, and gradient vector flow," Image Processing, IEEE Transactions

  • n 7, pp. 359-369, Mar 1998.
  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

slide-11
SLIDE 11

Segmentation Results

Left Maxillary Sinus Right Maxillary Sinus DR GVF DR GVF Front Back

slide-12
SLIDE 12

Segmentation Results

Errors (mm) as compared to hand segmented ground truth

slide-13
SLIDE 13

Segmentation Results

Errors (mm) as compared to hand segmented ground truth

slide-14
SLIDE 14

Statistical Shape Model (SSM)

[8][9]

PCA

[8] T. Cootes, C. Taylor, D. Cooper, and J. Graham, β€œActive shape models-their training and application,β€œ

Computer Vision and Image Understanding 61(1), pp. 38-59, 1995.

[9] G. Chintalapani, L. M. Ellingsen, O. Sadowsky, J. L. Prince, and R. H. Taylor, β€œStatistical atlases of bone

anatomy: construction, iterative improvement and validation," in Medical Image Computing and Computer- Assisted Intervention, pp. 499-506, 2007.

  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

slide-15
SLIDE 15

Statistical Shape Model (SSM)

Right Maxillary Sinus Left Maxillary Sinus Inferior Turbinate Middle Turbinate 1st Mode 2nd Mode

slide-16
SLIDE 16

Correspondence Improvement

[12]

βˆ—

[12] Seshamani S, Chintalapani G, Taylor RH, Iterative refinement of point correspondences for 3d statistical

shape models. MICCAI. 2011;417-425.

  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

slide-17
SLIDE 17

Correspondence Improvement

[12]

[12] Seshamani S, Chintalapani G, Taylor RH, Iterative refinement of point correspondences for 3d statistical

shape models. MICCAI. 2011;417-425.

  • 1. Automatic Segmentation

Template Creation Deformable Registration Segmentation Improvement

  • 2. Statistical Shape Modeling

Principal Component Analysis Correspondence Improvement

PCA

βˆ—

slide-18
SLIDE 18

Leave-one-out Analysis

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 10 20 30 40 50 60

Mean an Vertex Er Error

  • r (mm

mm) # # mo modes

Mid iddle Turbinate: Vertex Err Error

Iter_0 Iter_1 Iter_2 Iter_3

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 10 20 30 40 50 60

Residual al Sur urface Er Error

  • r (mm

mm) # # mo modes

Mid iddle Turbinate: Res esidual l Sur Surface Err Error

Iter_0 Iter_1 Iter_2 Iter_3

slide-19
SLIDE 19

Natural Variation

  • Hypothesis

is: Given CTs of π‘œ individuals, it is likely that the turbinates of each individual are at a different state in the nasal cycle than all others. Therefore, a statistical model of the turbinates built from these π‘œ CTs should also reflect natural variation.

slide-20
SLIDE 20

Natural Variation

  • Experiment
  • Built separate models for skull and inferior turbinates
  • We expect inferior turbinates to change, but the skull to not change.

This should be reflected in the mode weights when pre-op and post-

  • p models are projected onto our statistical model.

PATIENT X Pre-op CT Post-op CT 𝑐𝑗

𝐽1 = 𝑛𝑗 π‘ˆ π‘Š 𝐽1 βˆ’

π‘Š π‘Š

𝐽1 βˆ— =

π‘Š +

𝑗=1 π‘œπ‘‘

𝑐𝑗

𝐽1𝑛𝑗

𝑐𝑗

𝐽2 = 𝑛𝑗 π‘ˆ π‘Š 𝐽2 βˆ’

π‘Š π‘Š

𝐽2 βˆ— =

π‘Š +

𝑗=1 π‘œπ‘‘

𝑐𝑗

𝐽2𝑛𝑗

Compare mode weights

slide-21
SLIDE 21

Natural Variation

  • 4
  • 3
  • 2
  • 1

1 2 3 4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Mod

  • de weig

ights ts mo mode

Mode Weig ights: Bon Bone

P2a P2b

  • 4
  • 3
  • 2
  • 1

1 2 3 4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Mod

  • de weig

ights ts mo mode

Mode Weig ights: In Inferio ior Turbin inate

P2a P2b

slide-22
SLIDE 22

Natural Variation

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Dif ifference mo mode

Di Difference in n Mode Weig eights

Bone IT IT

slide-23
SLIDE 23

Natural Variation

slide-24
SLIDE 24

Natural Variation

slide-25
SLIDE 25

Population variation vs Natural Variation

slide-26
SLIDE 26

Summary

  • We have built an initial statistical shape model (SSM) of the paranasal

sinuses from CT scans of 53 different patients.

  • SSMs of erectile tissue in the sinuses reflect variations due to the nasal

cycle, which are captured in the modes of our PCA models.

  • A preliminary experiment with a single same-patient pre-op/post-op

CT image pair suggests that certain statistical modes are more sensitive than others in characterizing this variation.

slide-27
SLIDE 27

Future Work

  • We are currently working on constructing a larger statistical atlas of

the sinuses based on CT scans of 500 patients.

  • We hope to extend our exploration of the nasal cycle using a larger

number of same-patient longitudinal studies.

  • We are also working to incorporate our results into ongoing research
  • n intraoperative video-CT registration.
slide-28
SLIDE 28

References

[1] Hosemann W, Draf C. Danger points, complications and medico-legal aspects in endoscopic sinus surgery. GMS Current Topics in Otorhinolaryngology,

Head and Neck Surgery. 2013;12:Doc06.

[2] Hepworth EJ, Bucknor M, Patel A, Vaughan WC. Nationwide survey on the use of image-guided functional endoscopic sinus surgery. Otolaryngol Head

Neck Surg. 2006 Jul;135(1):68–73.

[3] Psaltis AJ, Soler ZM, Nguyen SA, Schlosser RJ. Changing trends in sinus and septal surgery, 2007 to 2009. Int Forum Allergy Rhinol. 2012 Sep-Oct;2(5):357–

361.

[4] Hasegawa M, Kern EB, The human nasal cycle. Mayo Clinic Proceedings. 1977 May;51:28-34 [5] Atanasov AT. Length of Periods in the Nasal Cycle during 24-Hours Registration. Open Journal of Biophysics. 2014;4:93-96 [6] Avants BB, Yushkevich P, Pluta J, Minko D, Korczykowski, M, Detre J, Gee JC. The optimal template effect in hippocampus studies of diseased populations.

NeuroImage;49(3):2457-2010.

[7] Avants BB, Tustison NJ, Song G, Cook PA, Klein A, and Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration.

  • NeuroImage. 2011;54(3):2033-2044. [8]Xu C, Prince JL. Gradient vector flow: A new external force for snakes. Computer Vision and Pattern Recognition, IEEE

Computer Society Conference on. 1997;66-71.

[8] Cootes T, Taylor C, Cooper D, Graham J. Active shape models-their training and application. Computer Vision and Image Understanding. 1995;61(1):38-59. [9] Chintalapani G, Ellingsen LM, Sadowsky O, Prince JL, and Taylor RH, Statistical atlases of bone anatomy: construction, iterative improvement and

  • validation. MICCAI. 2007;499-506.

[10] C. Xu and J. L. Prince, β€œGradient vector ow: A new external force for snakes," in Computer Vision and Pattern Recognition, IEEE Computer Society

Conference on, pp. 66-71, 1997.

[11] Xu C, Prince JL. Snakes, shapes, and gradient vector flow. Image Processing, IEEE Transactions on. 1998 Mar;7:359-369. [12] Seshamani S, Chintalapani G, Taylor RH, Iterative refinement of point correspondences for 3d statistical shape models. MICCAI. 2011;417-425. [13] Lorensen WE, Cline HE, Marching cubes: A high resolution 3d surface construction algorithm. SIGGRAPH. 1987;63-169. [14] Delgado-Gonzalo R,

Chenouard N, Unser M. Spline-based deforming ellipsoids for interactive 3D bioimage segmentation. Image Processing, IEEE Transactions on. 2013 Oct;22:3926-3940.

[14] Weiler K, Edge-based data structures for solid modeling in curved-surface environments. Computer Graphics and Applications, 1985 Jan;5:21-40.

slide-29
SLIDE 29

Thank you!

Questions?

Acknowledgement: This work is funded by NIH R01-EB015530: Enhanced Navigation for Endoscopic Sinus Surgery through Video Analysis.