TH‐C‐141‐01: Differentiation of Air From Bone Using MRI for Radiation Therapy and PET‐MRI Reconstruction

Shu-Hui Hsu, Y. Cao, J. Balter

Research output: Contribution to journalArticle

Abstract

Purpose: Differentiating air from bone is challenging with conventional MR images. While ultrashort TE (UTE) techniques have shown some promise, factors including noise, partial volume averaging, and imaging artifacts all may limit the discriminatory power of this technique. Methods: A biological phantom containing bone, bone marrow, solid tissue (with an air cavity), and fat was scanned with an 18‐channel surface coil on a 3T MR scanner (Siemens SKYRA). MRI volumes acquired included T1‐weighted gradient echo (TE 2.5 ms), T2‐weighted fast spin echo (TE 88 ms), and two echoes from a UTE acquisition (UTE1 TE 0.06 and UTE2 TE 4.46 ms). CT images for the phantom were acquired to define “true” air and bone regions using thresholds of −600 and 145, respectively. A range of intensity thresholds was used to test sensitivity in defining air regions in T1, T2, UTE1, UTE2 and difference of the UTE1 and UTE2 images. The number of voxels labeled as air in MR images and within air and bone regions defined in CT images were counted to calculate sensitivity and specificity. ROC curves were generated to evaluate bone and air discrimination in different MR image types. Results: ROC curves did not cross for the individual image types tested. UTE1 had the greatest discriminatory power (AUC 0.944), followed by UTE1–UTE2 (0.906), UTE2 (0.869), T1 (0.750), and T2 (0.651). UTE1 could not completely separate air from bone, generating false positive locations of air within the CT‐defined bone region of interest at sensitivities of 0.756 and higher for labeling air within the cavity defined on CT. Conclusion: While UTE1 reasonably separates air from bone, the persistence of false labeling of bone as air warrants further investigation. Subtraction of UTE1 and UTE2 did not improve sensitivity/specificity, indicating no added value of the 4.46 ms TE in separating bone from air.

Original languageEnglish (US)
Pages (from-to)539
Number of pages1
JournalMedical Physics
Volume40
Issue number6
DOIs
StatePublished - 2013
Externally publishedYes

Fingerprint

Radiotherapy
Air
Bone and Bones
ROC Curve
Sensitivity and Specificity
Artifacts
Area Under Curve
Noise
Bone Marrow
Fats

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

TH‐C‐141‐01 : Differentiation of Air From Bone Using MRI for Radiation Therapy and PET‐MRI Reconstruction. / Hsu, Shu-Hui; Cao, Y.; Balter, J.

In: Medical Physics, Vol. 40, No. 6, 2013, p. 539.

Research output: Contribution to journalArticle

@article{a41c046961784adcadbc69c6395ecb09,
title = "TH‐C‐141‐01: Differentiation of Air From Bone Using MRI for Radiation Therapy and PET‐MRI Reconstruction",
abstract = "Purpose: Differentiating air from bone is challenging with conventional MR images. While ultrashort TE (UTE) techniques have shown some promise, factors including noise, partial volume averaging, and imaging artifacts all may limit the discriminatory power of this technique. Methods: A biological phantom containing bone, bone marrow, solid tissue (with an air cavity), and fat was scanned with an 18‐channel surface coil on a 3T MR scanner (Siemens SKYRA). MRI volumes acquired included T1‐weighted gradient echo (TE 2.5 ms), T2‐weighted fast spin echo (TE 88 ms), and two echoes from a UTE acquisition (UTE1 TE 0.06 and UTE2 TE 4.46 ms). CT images for the phantom were acquired to define “true” air and bone regions using thresholds of −600 and 145, respectively. A range of intensity thresholds was used to test sensitivity in defining air regions in T1, T2, UTE1, UTE2 and difference of the UTE1 and UTE2 images. The number of voxels labeled as air in MR images and within air and bone regions defined in CT images were counted to calculate sensitivity and specificity. ROC curves were generated to evaluate bone and air discrimination in different MR image types. Results: ROC curves did not cross for the individual image types tested. UTE1 had the greatest discriminatory power (AUC 0.944), followed by UTE1–UTE2 (0.906), UTE2 (0.869), T1 (0.750), and T2 (0.651). UTE1 could not completely separate air from bone, generating false positive locations of air within the CT‐defined bone region of interest at sensitivities of 0.756 and higher for labeling air within the cavity defined on CT. Conclusion: While UTE1 reasonably separates air from bone, the persistence of false labeling of bone as air warrants further investigation. Subtraction of UTE1 and UTE2 did not improve sensitivity/specificity, indicating no added value of the 4.46 ms TE in separating bone from air.",
author = "Shu-Hui Hsu and Y. Cao and J. Balter",
year = "2013",
doi = "10.1118/1.4815769",
language = "English (US)",
volume = "40",
pages = "539",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "6",

}

TY - JOUR

T1 - TH‐C‐141‐01

T2 - Differentiation of Air From Bone Using MRI for Radiation Therapy and PET‐MRI Reconstruction

AU - Hsu, Shu-Hui

AU - Cao, Y.

AU - Balter, J.

PY - 2013

Y1 - 2013

N2 - Purpose: Differentiating air from bone is challenging with conventional MR images. While ultrashort TE (UTE) techniques have shown some promise, factors including noise, partial volume averaging, and imaging artifacts all may limit the discriminatory power of this technique. Methods: A biological phantom containing bone, bone marrow, solid tissue (with an air cavity), and fat was scanned with an 18‐channel surface coil on a 3T MR scanner (Siemens SKYRA). MRI volumes acquired included T1‐weighted gradient echo (TE 2.5 ms), T2‐weighted fast spin echo (TE 88 ms), and two echoes from a UTE acquisition (UTE1 TE 0.06 and UTE2 TE 4.46 ms). CT images for the phantom were acquired to define “true” air and bone regions using thresholds of −600 and 145, respectively. A range of intensity thresholds was used to test sensitivity in defining air regions in T1, T2, UTE1, UTE2 and difference of the UTE1 and UTE2 images. The number of voxels labeled as air in MR images and within air and bone regions defined in CT images were counted to calculate sensitivity and specificity. ROC curves were generated to evaluate bone and air discrimination in different MR image types. Results: ROC curves did not cross for the individual image types tested. UTE1 had the greatest discriminatory power (AUC 0.944), followed by UTE1–UTE2 (0.906), UTE2 (0.869), T1 (0.750), and T2 (0.651). UTE1 could not completely separate air from bone, generating false positive locations of air within the CT‐defined bone region of interest at sensitivities of 0.756 and higher for labeling air within the cavity defined on CT. Conclusion: While UTE1 reasonably separates air from bone, the persistence of false labeling of bone as air warrants further investigation. Subtraction of UTE1 and UTE2 did not improve sensitivity/specificity, indicating no added value of the 4.46 ms TE in separating bone from air.

AB - Purpose: Differentiating air from bone is challenging with conventional MR images. While ultrashort TE (UTE) techniques have shown some promise, factors including noise, partial volume averaging, and imaging artifacts all may limit the discriminatory power of this technique. Methods: A biological phantom containing bone, bone marrow, solid tissue (with an air cavity), and fat was scanned with an 18‐channel surface coil on a 3T MR scanner (Siemens SKYRA). MRI volumes acquired included T1‐weighted gradient echo (TE 2.5 ms), T2‐weighted fast spin echo (TE 88 ms), and two echoes from a UTE acquisition (UTE1 TE 0.06 and UTE2 TE 4.46 ms). CT images for the phantom were acquired to define “true” air and bone regions using thresholds of −600 and 145, respectively. A range of intensity thresholds was used to test sensitivity in defining air regions in T1, T2, UTE1, UTE2 and difference of the UTE1 and UTE2 images. The number of voxels labeled as air in MR images and within air and bone regions defined in CT images were counted to calculate sensitivity and specificity. ROC curves were generated to evaluate bone and air discrimination in different MR image types. Results: ROC curves did not cross for the individual image types tested. UTE1 had the greatest discriminatory power (AUC 0.944), followed by UTE1–UTE2 (0.906), UTE2 (0.869), T1 (0.750), and T2 (0.651). UTE1 could not completely separate air from bone, generating false positive locations of air within the CT‐defined bone region of interest at sensitivities of 0.756 and higher for labeling air within the cavity defined on CT. Conclusion: While UTE1 reasonably separates air from bone, the persistence of false labeling of bone as air warrants further investigation. Subtraction of UTE1 and UTE2 did not improve sensitivity/specificity, indicating no added value of the 4.46 ms TE in separating bone from air.

UR - http://www.scopus.com/inward/record.url?scp=85024815429&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85024815429&partnerID=8YFLogxK

U2 - 10.1118/1.4815769

DO - 10.1118/1.4815769

M3 - Article

AN - SCOPUS:85024815429

VL - 40

SP - 539

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 6

ER -