Imaging and quantification of endothelial cell loss in eye bank prepared DMEK grafts using trainable segmentation software

Griffin J. Jardine, Jeffrey D. Holiman, Christopher G. Stoeger, Winston Chamberlain

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

Purpose: To improve accuracy and efficiency in quantifying the endothelial cell loss (ECL) in eye bank preparation of corneal endothelial grafts. Methods: Eight cadaveric corneas were subjected to Descemet Membrane Endothelial Keratoplasty (DMEK) preparation. The endothelial surfaces were stained with a viability stain, calcein AM dye (CAM) and then captured by a digital camera. The ECL rates were quantified in these images by three separate readers using trainable segmentation, a plug-in feature from the imaging software, Fiji. Images were also analyzed by Adobe Photoshop for comparison. Mean times required to process the images were measured between the two modalities. Results: The mean ECL (with standard deviation) as analyzed by Fiji was 22.5% (6.5%) and Adobe was 18.7% (7.0%; p=0.04). The mean time required to process the images through the two different imaging methods was 19.9min (7.5) for Fiji and 23.4min (12.9) for Adobe (p=0.17). Conclusions: Establishing an accurate, efficient and reproducible means of quantifying ECL in graft preparation and surgical techniques can provide insight to the safety, long-term potential of the graft tissues as well as provide a quality control measure for eye banks and surgeons. Trainable segmentation in Fiji software using CAM is a novel approach to measuring ECL that captured a statistically significantly higher percentage of ECL comparable to Adobe and was more accurate in standardized testing. Interestingly, ECL as determined using both methods in eye bank-prepared DMEK grafts exceeded 18% on average.

Original languageEnglish (US)
Pages (from-to)894-901
Number of pages8
JournalCurrent Eye Research
Volume39
Issue number9
DOIs
StatePublished - 2014

Fingerprint

Eye Banks
Descemet Membrane
Corneal Transplantation
Fiji
Software
Endothelial Cells
Transplants
Coloring Agents
Quality Control
Cornea
Safety

Keywords

  • Corneal transplant surgery
  • Descemet Membrane Endothelial Keratoplasty
  • Endothelial cell loss
  • Eye bank
  • Trainable segmentation

ASJC Scopus subject areas

  • Ophthalmology
  • Sensory Systems
  • Cellular and Molecular Neuroscience
  • Medicine(all)

Cite this

Imaging and quantification of endothelial cell loss in eye bank prepared DMEK grafts using trainable segmentation software. / Jardine, Griffin J.; Holiman, Jeffrey D.; Stoeger, Christopher G.; Chamberlain, Winston.

In: Current Eye Research, Vol. 39, No. 9, 2014, p. 894-901.

Research output: Contribution to journalArticle

Jardine, Griffin J. ; Holiman, Jeffrey D. ; Stoeger, Christopher G. ; Chamberlain, Winston. / Imaging and quantification of endothelial cell loss in eye bank prepared DMEK grafts using trainable segmentation software. In: Current Eye Research. 2014 ; Vol. 39, No. 9. pp. 894-901.
@article{72716b6401ce408492df24a9ff30a873,
title = "Imaging and quantification of endothelial cell loss in eye bank prepared DMEK grafts using trainable segmentation software",
abstract = "Purpose: To improve accuracy and efficiency in quantifying the endothelial cell loss (ECL) in eye bank preparation of corneal endothelial grafts. Methods: Eight cadaveric corneas were subjected to Descemet Membrane Endothelial Keratoplasty (DMEK) preparation. The endothelial surfaces were stained with a viability stain, calcein AM dye (CAM) and then captured by a digital camera. The ECL rates were quantified in these images by three separate readers using trainable segmentation, a plug-in feature from the imaging software, Fiji. Images were also analyzed by Adobe Photoshop for comparison. Mean times required to process the images were measured between the two modalities. Results: The mean ECL (with standard deviation) as analyzed by Fiji was 22.5{\%} (6.5{\%}) and Adobe was 18.7{\%} (7.0{\%}; p=0.04). The mean time required to process the images through the two different imaging methods was 19.9min (7.5) for Fiji and 23.4min (12.9) for Adobe (p=0.17). Conclusions: Establishing an accurate, efficient and reproducible means of quantifying ECL in graft preparation and surgical techniques can provide insight to the safety, long-term potential of the graft tissues as well as provide a quality control measure for eye banks and surgeons. Trainable segmentation in Fiji software using CAM is a novel approach to measuring ECL that captured a statistically significantly higher percentage of ECL comparable to Adobe and was more accurate in standardized testing. Interestingly, ECL as determined using both methods in eye bank-prepared DMEK grafts exceeded 18{\%} on average.",
keywords = "Corneal transplant surgery, Descemet Membrane Endothelial Keratoplasty, Endothelial cell loss, Eye bank, Trainable segmentation",
author = "Jardine, {Griffin J.} and Holiman, {Jeffrey D.} and Stoeger, {Christopher G.} and Winston Chamberlain",
year = "2014",
doi = "10.3109/02713683.2014.887120",
language = "English (US)",
volume = "39",
pages = "894--901",
journal = "Current Eye Research",
issn = "0271-3683",
publisher = "Informa Healthcare",
number = "9",

}

TY - JOUR

T1 - Imaging and quantification of endothelial cell loss in eye bank prepared DMEK grafts using trainable segmentation software

AU - Jardine, Griffin J.

AU - Holiman, Jeffrey D.

AU - Stoeger, Christopher G.

AU - Chamberlain, Winston

PY - 2014

Y1 - 2014

N2 - Purpose: To improve accuracy and efficiency in quantifying the endothelial cell loss (ECL) in eye bank preparation of corneal endothelial grafts. Methods: Eight cadaveric corneas were subjected to Descemet Membrane Endothelial Keratoplasty (DMEK) preparation. The endothelial surfaces were stained with a viability stain, calcein AM dye (CAM) and then captured by a digital camera. The ECL rates were quantified in these images by three separate readers using trainable segmentation, a plug-in feature from the imaging software, Fiji. Images were also analyzed by Adobe Photoshop for comparison. Mean times required to process the images were measured between the two modalities. Results: The mean ECL (with standard deviation) as analyzed by Fiji was 22.5% (6.5%) and Adobe was 18.7% (7.0%; p=0.04). The mean time required to process the images through the two different imaging methods was 19.9min (7.5) for Fiji and 23.4min (12.9) for Adobe (p=0.17). Conclusions: Establishing an accurate, efficient and reproducible means of quantifying ECL in graft preparation and surgical techniques can provide insight to the safety, long-term potential of the graft tissues as well as provide a quality control measure for eye banks and surgeons. Trainable segmentation in Fiji software using CAM is a novel approach to measuring ECL that captured a statistically significantly higher percentage of ECL comparable to Adobe and was more accurate in standardized testing. Interestingly, ECL as determined using both methods in eye bank-prepared DMEK grafts exceeded 18% on average.

AB - Purpose: To improve accuracy and efficiency in quantifying the endothelial cell loss (ECL) in eye bank preparation of corneal endothelial grafts. Methods: Eight cadaveric corneas were subjected to Descemet Membrane Endothelial Keratoplasty (DMEK) preparation. The endothelial surfaces were stained with a viability stain, calcein AM dye (CAM) and then captured by a digital camera. The ECL rates were quantified in these images by three separate readers using trainable segmentation, a plug-in feature from the imaging software, Fiji. Images were also analyzed by Adobe Photoshop for comparison. Mean times required to process the images were measured between the two modalities. Results: The mean ECL (with standard deviation) as analyzed by Fiji was 22.5% (6.5%) and Adobe was 18.7% (7.0%; p=0.04). The mean time required to process the images through the two different imaging methods was 19.9min (7.5) for Fiji and 23.4min (12.9) for Adobe (p=0.17). Conclusions: Establishing an accurate, efficient and reproducible means of quantifying ECL in graft preparation and surgical techniques can provide insight to the safety, long-term potential of the graft tissues as well as provide a quality control measure for eye banks and surgeons. Trainable segmentation in Fiji software using CAM is a novel approach to measuring ECL that captured a statistically significantly higher percentage of ECL comparable to Adobe and was more accurate in standardized testing. Interestingly, ECL as determined using both methods in eye bank-prepared DMEK grafts exceeded 18% on average.

KW - Corneal transplant surgery

KW - Descemet Membrane Endothelial Keratoplasty

KW - Endothelial cell loss

KW - Eye bank

KW - Trainable segmentation

UR - http://www.scopus.com/inward/record.url?scp=84906282062&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84906282062&partnerID=8YFLogxK

U2 - 10.3109/02713683.2014.887120

DO - 10.3109/02713683.2014.887120

M3 - Article

VL - 39

SP - 894

EP - 901

JO - Current Eye Research

JF - Current Eye Research

SN - 0271-3683

IS - 9

ER -