TY - JOUR
T1 - Objective assessment of stored blood quality by deep learning
AU - Doan, Minh
AU - Sebastian, Joseph A.
AU - Caicedo, Juan C.
AU - Siegert, Stefanie
AU - Roch, Aline
AU - Turner, Tracey R.
AU - Mykhailova, Olga
AU - Pinto, Ruben N.
AU - McQuin, Claire
AU - Goodman, Allen
AU - Parsons, Michael J.
AU - Wolkenhauer, Olaf
AU - Hennig, Holger
AU - Singh, Shantanu
AU - Wilson, Anne
AU - Acker, Jason P.
AU - Rees, Paul
AU - Kolios, Michael C.
AU - Carpenter, Anne E.
AU - Geman, Donald
N1 - Funding Information:
via a Collaborative Health Research Projects Grant 315271 “Characterization of blood storage lesions using photoacoustic technologies” (to M.C.K. and J.P.A.); and a grant administered by Carigest S.A. of Geneva, Switzerland (to A.R.). The Canadian Blood Services research program is funded by the federal (Health Canada), provincial, and territorial Ministries of Health. Experiments were performed at the University of Alberta Faculty of Medicine & Dentistry Flow Cytometry Facility, which receives financial support from the Faculty of Medicine & Dentistry and Canada Foundation for Innovation awards to contributing investigators. The views expressed herein do not represent the views of the Canadian federal government or any other funding agencies.
Funding Information:
We thank the staff of the netCAD Blood for Research Facility, Centre for Innovation, Canadian Blood Services, Sophie Waldvogel, and all the staff at the Transfusion Center of the University Hospital of Geneva (Switzerland) for providing blood samples and quality-control data, and the generosity of the blood donors who made this research possible; T. C. Chang for consultations associated with validating the selection of images for the truth populations used for analysis, and for the development of the red blood cell gating and filtering template on the IDEAS software platform; The Lunenfeld Tanenbaum Research Institute flow cytometry facility for providing access for image flow cytometry experiments (supported through grants from the Canada Foundation For Innovation); M. H. Rohban for his expert consultations on developing fundamental concepts and critical elements of the machine-learning and deep-learning frameworks throughout the study; and Maren Buettner for critical feedback on the manuscript. Funding for this project was provided by US National Science Foundation/UK Biotechnology and Biological Sciences Research Council Joint Grant NSF DBI 1458626 and BB/N005163 (to A.E.C. and P.R.); Biotechnology and Biological Sciences Research Council Grant BB/P026818/1 (to P.R.); Natural Sciences and Engineering Research Council of Canada and the Canadian Institutes of Health Research, via a Collaborative Health Research Projects Grant 315271 ?Characterization of blood storage lesions using photoacoustic technologies? (to M.C.K. and J.P.A.); and a grant administered by Carigest S.A. of Geneva, Switzerland (to A.R.). The Canadian Blood Services research program is funded by the federal (Health Canada), provincial, and territorial Ministries of Health. Experiments were performed at the University of Alberta Faculty of Medicine & Dentistry Flow Cytometry Facility, which receives financial support from the Faculty of Medicine & Dentistry and Canada Foundation for Innovation awards to contributing investigators. The views expressed herein do not represent the views of the Canadian federal government or any other funding agencies.
Funding Information:
platform; The Lunenfeld Tanenbaum Research Institute flow cytometry facility for providing access for image flow cytometry experiments (supported through grants from the Canada Foundation For Innovation); M. H. Rohban for his expert consultations on developing fundamental concepts and critical elements of the machine-learning and deep-learning frameworks throughout the study; and Maren Buettner for critical feedback on the manuscript. Funding for this project was provided by US National Science Foundation/UK Biotechnology and Biological Sciences Research Council Joint Grant NSF DBI 1458626 and BB/N005163 (to A.E.C. and P.R.); Biotechnology and Biological Sciences Research Council Grant BB/P026818/1 (to P.R.); Natural Sciences and Engineering Research Council of Canada and the Canadian Institutes of Health Research,
Publisher Copyright:
© 2020 National Academy of Sciences. All rights reserved.
PY - 2020/9/1
Y1 - 2020/9/1
N2 - Stored red blood cells (RBCs) are needed for life-saving blood transfusions, but they undergo continuous degradation. RBC storage lesions are often assessed by microscopic examination or biochemical and biophysical assays, which are complex, time-consuming, and destructive to fragile cells. Here we demonstrate the use of label-free imaging flow cytometry and deep learning to characterize RBC lesions. Using brightfield images, a trained neural network achieved 76.7% agreement with experts in classifying seven clinically relevant RBC morphologies associated with storage lesions, comparable to 82.5% agreement between different experts. Given that human observation and classification may not optimally discern RBC quality, we went further and eliminated subjective human annotation in the training step by training a weakly supervised neural network using only storage duration times. The feature space extracted by this network revealed a chronological progression of morphological changes that better predicted blood quality, as measured by physiological hemolytic assay readouts, than the conventional expert-assessed morphology classification system. With further training and clinical testing across multiple sites, protocols, and instruments, deep learning and label-free imaging flow cytometry might be used to routinely and objectively assess RBC storage lesions. This would automate a complex protocol, minimize laboratory sample handling and preparation, and reduce the impact of procedural errors and discrepancies between facilities and blood donors. The chronology-based machine-learning approach may also improve upon humans’ assessment of morphological changes in other biomedically important progressions, such as differentiation and metastasis.
AB - Stored red blood cells (RBCs) are needed for life-saving blood transfusions, but they undergo continuous degradation. RBC storage lesions are often assessed by microscopic examination or biochemical and biophysical assays, which are complex, time-consuming, and destructive to fragile cells. Here we demonstrate the use of label-free imaging flow cytometry and deep learning to characterize RBC lesions. Using brightfield images, a trained neural network achieved 76.7% agreement with experts in classifying seven clinically relevant RBC morphologies associated with storage lesions, comparable to 82.5% agreement between different experts. Given that human observation and classification may not optimally discern RBC quality, we went further and eliminated subjective human annotation in the training step by training a weakly supervised neural network using only storage duration times. The feature space extracted by this network revealed a chronological progression of morphological changes that better predicted blood quality, as measured by physiological hemolytic assay readouts, than the conventional expert-assessed morphology classification system. With further training and clinical testing across multiple sites, protocols, and instruments, deep learning and label-free imaging flow cytometry might be used to routinely and objectively assess RBC storage lesions. This would automate a complex protocol, minimize laboratory sample handling and preparation, and reduce the impact of procedural errors and discrepancies between facilities and blood donors. The chronology-based machine-learning approach may also improve upon humans’ assessment of morphological changes in other biomedically important progressions, such as differentiation and metastasis.
KW - Cell morphology
KW - Deep learning
KW - Stored blood quality
KW - Weakly supervised learning
UR - http://www.scopus.com/inward/record.url?scp=85090511449&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85090511449&partnerID=8YFLogxK
U2 - 10.1073/pnas.2001227117
DO - 10.1073/pnas.2001227117
M3 - Article
C2 - 32839303
AN - SCOPUS:85090511449
VL - 117
SP - 21381
EP - 21390
JO - Proceedings of the National Academy of Sciences of the United States of America
JF - Proceedings of the National Academy of Sciences of the United States of America
SN - 0027-8424
IS - 35
ER -