Computerized segmentation algorithm with personalized atlases of murine MRIs in a SV40 Large T-antigen mouse mammary cancer model

Adam R. Sibley, Erica Markiewicz, Devkumar Mustafi, Xiaobing Fan, Suzanne Conzen, Greg Karczmar, Maryellen L. Giger

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Quantities of MRI data, much larger than can be objectively and efficiently analyzed manually, are routinely generated in preclinical research. We aim to develop an automated image segmentation and registration pipeline to aid in analysis of image data from our high-throughput 9.4 Tesla small animal MRI imaging center. T2-weighted, fat-suppressed MRIs were acquired over 4 life-cycle time-points [up to 12 to 18 weeks] of twelve C3(1) SV40 Large T-antigen mice for a total of 46 T2-weighted MRI volumes; each with a matrix size of 192 x 256, 62 slices, in plane resolution 0.1 mm, and slice thickness 0.5 mm. These image sets were acquired with the goal of tracking and quantifying progression of mammary intraepithelial neoplasia (MIN) to invasive cancer in mice, believed to be similar to ductal carcinoma in situ (DCIS) in humans. Our segmentation algorithm takes 2D seed-points drawn by the user at the center of the 4 coregistered volumes associated with each mouse. The level set then evolves in 3D from these 2D seeds. The contour evolution incorporates texture information, edge information, and a statistical shape model in a two-step process. Volumetric DICE coefficients comparing the automatic with manual segmentations were computed and ranged between 0.75 and 0.58 for averages over the 4 life-cycle time points of the mice. Incorporation of these personalized atlases with intra and inter mouse registration is expected to enable locally and globally tracking of the morphological and textural changes in the mammary tissue and associated lesions of these mice.

Original languageEnglish (US)
Title of host publicationMedical Imaging 2016
Subtitle of host publicationBiomedical Applications in Molecular, Structural, and Functional Imaging
EditorsBarjor Gimi, Andrzej Krol
PublisherSPIE
ISBN (Electronic)9781510600232
DOIs
StatePublished - Jan 1 2016
Externally publishedYes
EventMedical Imaging 2016: Biomedical Applications in Molecular, Structural, and Functional Imaging - San Diego, United States
Duration: Mar 1 2016Mar 3 2016

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume9788
ISSN (Print)1605-7422

Other

OtherMedical Imaging 2016: Biomedical Applications in Molecular, Structural, and Functional Imaging
CountryUnited States
CitySan Diego
Period3/1/163/3/16

    Fingerprint

Keywords

  • Breast cancer
  • Ductal carcinoma in situ
  • Image analysis
  • Mammary intraepithelial neoplasia
  • MRI
  • Murine mammary gland
  • Preclinical
  • Radiomics

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Biomaterials
  • Atomic and Molecular Physics, and Optics
  • Radiology Nuclear Medicine and imaging

Cite this

Sibley, A. R., Markiewicz, E., Mustafi, D., Fan, X., Conzen, S., Karczmar, G., & Giger, M. L. (2016). Computerized segmentation algorithm with personalized atlases of murine MRIs in a SV40 Large T-antigen mouse mammary cancer model. In B. Gimi, & A. Krol (Eds.), Medical Imaging 2016: Biomedical Applications in Molecular, Structural, and Functional Imaging [97882M] (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 9788). SPIE. https://doi.org/10.1117/12.2217425