TY - JOUR
T1 - Artificial intelligence-assisted identification and quantification of osteoclasts
AU - Emmanuel, Thomas
AU - Brüel, Annemarie
AU - Thomsen, Jesper Skovhus
AU - Steiniche, Torben
AU - Brent, Mikkel Bo
PY - 2021/1
Y1 - 2021/1
N2 - Quantification of osteoclasts to assess bone resorption is a time-consuming and tedious process. Since the inception of bone histomorphometry and manual counting of osteoclasts using bright-field microscopy, several approaches have been proposed to accelerate the counting process using both free and commercially available software. However, most of the present alternatives depend on manual or semi-automatic color segmentation and do not take advantage of artificial intelligence (AI). The present study directly compare estimates of osteoclast-covered surfaces (Oc.S/BS) obtained by the conventional manual method using a bright-field microscope to that obtained by a new AI-assisted method. We present a detailed step-by-step guide for the AI-based method. Tibiae from Wistar rats were either enzymatically stained for TRAP or immunostained for cathepsin K to identify osteoclasts. We found that estimation of Oc.S/BS by the new AI-assisted method was considerably less time-consuming, while still providing similar results to the conventional manual method. In addition, the retrainable AI-module used in the present study allows for fully automated overnight batch processing of multiple annotated sections.
AB - Quantification of osteoclasts to assess bone resorption is a time-consuming and tedious process. Since the inception of bone histomorphometry and manual counting of osteoclasts using bright-field microscopy, several approaches have been proposed to accelerate the counting process using both free and commercially available software. However, most of the present alternatives depend on manual or semi-automatic color segmentation and do not take advantage of artificial intelligence (AI). The present study directly compare estimates of osteoclast-covered surfaces (Oc.S/BS) obtained by the conventional manual method using a bright-field microscope to that obtained by a new AI-assisted method. We present a detailed step-by-step guide for the AI-based method. Tibiae from Wistar rats were either enzymatically stained for TRAP or immunostained for cathepsin K to identify osteoclasts. We found that estimation of Oc.S/BS by the new AI-assisted method was considerably less time-consuming, while still providing similar results to the conventional manual method. In addition, the retrainable AI-module used in the present study allows for fully automated overnight batch processing of multiple annotated sections.
KW - AI-assisted image processing
KW - Artificial intelligence-assisted identification and quantification of osteoclasts
KW - Bone histomorphometry
KW - Osteoclasts
UR - https://www.scopus.com/pages/publications/85101516262
U2 - 10.1016/j.mex.2021.101272
DO - 10.1016/j.mex.2021.101272
M3 - Journal article
C2 - 34434793
SN - 2215-0161
VL - 8
JO - MethodsX
JF - MethodsX
M1 - 101272
ER -