Numerical Model Development Part I – Pediatric Anatomy Albert I. King Bioengineering Center

1. 2006 SCIENTIFIC SYMPOSIUM OF THE SCIB The Digital Child Project Numerical Model Development Part I – Pediatric Anatomy Albert I. King Bioengineering Center Wayne State University Birmingham, Alabama December 13, 2006

2. Objectives • Identify level of detail needed for pediatric FE models, focusing on anatomical differences between adults and children • Create appropriate whole body anthropometry for 3, 6, and 10 year olds using clinical CT and MRI data

3. Pediatric Anatomy • Child ≠ Small Adult • Anatomical and morphological changes throughout maturation • Example: • Ossification of skeletal structures • Skull • Spine

4. Pediatric Anatomy NOT TO SCALE 2 YO 5 YO ADULT Images from http://www.boneclones.com

5. Pediatric Anatomy Image from http://faculty.washington.edu/chudler/dev.html

6. Pediatric Anatomy NOT TO SCALE Images from http://www.yoursurgery.com/ProcedureDetails.cfm?BR=4&Proc=19

7. Pediatric Anatomy NOT TO SCALE Images from http://faculty.clintoncc.suny.edu/faculty/Michael.Gregory/files/ Bio%20102/Bio%20102%20lectures/Motor%20Systems/infant_skull.jpg and http://www.uoftbookstore.com/online/prodimg/49090.jpg

8. Pediatric Anatomy ADULT 6 YO 3 YO INFANT NOT TO SCALE Image altered from Kumaresan, et al. (2000)

9. Procure CT and MRI images from medical record (body parts from different children) Create 3D surface Scale to average size, combine to create whole body if needed Process images to improve image quality, segmentation of structures Create whole body CT and MRI images from pediatric cadavers Position in appropriate seating posture Flowchart

10. Timeline

11. HIC Approval • Institutional Review Board: Human Investigation Committee of Wayne State University • Approved collection of pediatric image data from medical records on September 1, 2006 • Application in process to obtain whole body images from pediatric cadavers (nondestructive)

12. Data Collection • MRI data collected from 11 subjects, further MRI and CT data in progress in collaboration with the Chair of the Radiology Dept. • Target ages (either gender) • 2.5-3.5 years • 5.5-6.5 years • 9.5-10.5 years

13. Data Collection • MRI data for each subject includes: • Several orientations • Axial, coronal, sagittal • Several types • T1, T2, etc. • May include contrast • Some MRAs (blood vessels)

14. Segmentation • Challenges in image processing: • Abnormal anatomy • Slice orientation • Image quality • Slice thickness and resolution • Noise • Incomplete structures of interest

15. Abnormal anatomy present- reason for scans

16. Segmentation • Solutions - Abnormal anatomy • Bilateral structures can be reflected • Removal of pathological tissue formations and manual interpolation • If too abnormal, collect new data from medical record • Pediatric cadavers chosen may have limited abnormalities if death due to trauma

17. Slice orientation- non-orthogonal

18. Segmentation • Solutions – Slice orientation • Can make structure identification difficult for the engineer, consult radiologist • Translation and/or rotation of structures may be necessary for integration into whole body model (addressed later)

19. Image Quality 3D Doctor E-Film Lite 3D Slicer Mimics

20. Slice thickness- ranges from 1.8-5.0 mm, usually 4-5 mm

21. Segmentation • Solutions – Slice thickness and resolution • Each subject has numerous image series, choose the highest resolution • Smoothing after surface generated • If pediatric cadavers become available, we can choose finer cuts and higher resolutions with no radiation exposure hazard

22. Noise

23. Segmentation using Mimics without noise removal Segmentation with noise removed

24. Segmentation • Solutions – Noise • Manual removal with image processing and medical image segmentation software packages • May not affect structure of interest • Collect new data from the medical record

25. Some structures are incomplete due to Field of View In this case, the brain Often a problem in junction between thorax and abdomen (liver often incomplete in both)

26. Segmentation • Solutions- Incomplete structures • Manual interpolation • Collect new data for that body part from medical record • Could be solved if pediatric cadavers become available

27. Scaling • Combine body parts from different subjects to create a whole body model which is “average” sized • Preserve developmental anatomy • Scaling child to child of similar maturation level (NOT adult to child) • Geometric scaling only (based on Irwin, et al. 1997)

28. Scaling • “Average” child anthropometry by age determined from Snyder, et al. (1977) • External measurements only, no data available on relationship between external landmarks and internal landmarks (noted by Reed, et al. at UMTRI, 2005, 2006)

29. Scaling • Snyder data provided • Heights • Breadths • Circumferences of external landmarks • Age ranges of interest: • 2.0-3.5, 5.5-6.5, and 9.5-10.5 years

30. Head Dimensions in the sagittal and coronal planes Image altered from http://ovrt.nist.gov/projects/anthrokids/

31. Thorax Dimensions in the coronal plane Sagittal plane data also available Image altered from http://ovrt.nist.gov/projects/anthrokids/

32. Scaling • Challenges • All anthropometric data cannot be applied to each subject’s scans • For example, this abdomen MRI precludes measuring chest breadth at axilla Breadth at Axilla Natural Waist Breadth Waist Breadth

33. Scaling • Solutions- Scaling • Scale using all available external measurements • When finished, check realism of complete body in terms of proportion • Use whole body images from pediatric cadavers to determine relationships between internal and external geometries

34. Positioning • Complete 3D solid model must be properly positioned for use in analysis • Translation of entire model • Rotation of joints • Automotive seating postures to be based on Reed, et al. (2005, 2006) • Data not available for three year olds

35. Positioning • Challenge • Snyder’s seated anthropometries not representative of automotive seating postures • Sources of such seating postures for children incomplete Image altered from http://ovrt.nist.gov/projects/anthrokids/

36. Positioning Positions of the extremities not reported Data available for 6 YO and 10 YO, including head angle and positions relative to H-point Image altered from Reed, et al. (2005)

37. Positioning • Solutions- Positioning • Extrapolation • From Reed, et al. data • May perform small study at WSU • Will not be representative of the population

38. Proposed Positioning Study

39. Discussion and Conclusions • In creating accurate pediatric 3D model surfaces: • Pediatric medical images of a lower quality than adult due to standard procedures • Issues can be overcome will the use of a combination of software and manual editing

40. Discussion and Conclusions • In the absence of an average sized pediatric cadaver, geometric scaling between subjects of similar maturation levels will be necessary • Positioning of the 3D model involves sparse data sets, but reasonable postures can be achieved with additional experimental information

41. Future Work • Collect more MRI and CT data, especially CT in regions of partial ossification • Continue work to procure whole body images of pediatric cadavers • Determine appropriate scaling method and calculate component locations (joint centers)

42. Future Work • Evaluate open source and commercial image processing/segmentation software • Evaluate different segmentation methods using known partially ossified structures • Create segmentation protocol appropriate for pediatric anatomy

43. Thank You

45. INFANT 1-3 YO 3-6 YO 11-14 YO ADULT Image altered from Yoganandan, et al. (1999)

47. Pelvic Dimensions in the coronal plane Image altered from http://ovrt.nist.gov/projects/anthrokids/

48. Positioning • Automotive seating postures to be based on Reed, et al. (2005, 2006) • Data not available for three year old Average age = 8.4 years