TY - GEN
T1 - Using hidden Markov models to determine changes in subject data over time, studying the immunoregulatory effect of mesenchymal stem cells
AU - Black, Edgar F.
AU - Marini, Luigi
AU - Vaidya, Ashwini
AU - Berman, Dora
AU - Willman, Melissa
AU - Salomon, Dan
AU - Bartholomew, Amelia
AU - Kenyon, Norma
AU - McHenry, Kenton
N1 - Publisher Copyright:
© 2014 IEEE.
PY - 2014/12/2
Y1 - 2014/12/2
N2 - A novel application of Hidden Markov Models is used to help research intended to test the immunuregulatory effects of mesenchymal stem cells in a cynomolgus monkey model of islet transplantation. The Hidden Markov Model, an unsupervised learning data mining technique, is used to automatically determine the postoperative day (POD) corresponding to a decrease of graft function, a possible sign of transplant rejection, on nonhuman primates after isolated islet cell transplant. Currently, decrease of graft function is being determined solely on experts' judgment. Further, information gathered from the evaluation of constructed Hidden Markov Models is used as part of a clustering method to aggregate the nonhuman subjects into groups or clusters with the objective of finding similarities that could potentially help predict the health outcome of subjects undergoing postoperative care. Results on expert labelled data show the HMM to be accurate 60% of the time. Clusters based on the HMMs further suggest a possible correspondence between donor haplotypes matching and loss of function outcomes.
AB - A novel application of Hidden Markov Models is used to help research intended to test the immunuregulatory effects of mesenchymal stem cells in a cynomolgus monkey model of islet transplantation. The Hidden Markov Model, an unsupervised learning data mining technique, is used to automatically determine the postoperative day (POD) corresponding to a decrease of graft function, a possible sign of transplant rejection, on nonhuman primates after isolated islet cell transplant. Currently, decrease of graft function is being determined solely on experts' judgment. Further, information gathered from the evaluation of constructed Hidden Markov Models is used as part of a clustering method to aggregate the nonhuman subjects into groups or clusters with the objective of finding similarities that could potentially help predict the health outcome of subjects undergoing postoperative care. Results on expert labelled data show the HMM to be accurate 60% of the time. Clusters based on the HMMs further suggest a possible correspondence between donor haplotypes matching and loss of function outcomes.
KW - Hidden Markov Models
KW - data mining
KW - unsupervised learning
UR - http://www.scopus.com/inward/record.url?scp=84919482451&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84919482451&partnerID=8YFLogxK
U2 - 10.1109/eScience.2014.29
DO - 10.1109/eScience.2014.29
M3 - Conference contribution
C2 - 26075290
AN - SCOPUS:84919482451
T3 - Proceedings - 2014 IEEE 10th International Conference on eScience, eScience 2014
SP - 83
EP - 91
BT - Proceedings - 2014 IEEE 10th International Conference on eScience, eScience 2014
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 10th IEEE International Conference on eScience, eScience 2014
Y2 - 20 October 2014 through 24 October 2014
ER -