Purpose
Supervised machine learning (ML) provides a compelling alternative to traditional model fitting for parameter mapping in quantitative MRI. The aim of this work is to demonstrate and quantify the effect of different training data distributions on the accuracy and precision of parameter estimates when supervised ML is used for fitting.
Methods
We fit a two‐ and three‐compartment biophysical model to diffusion measurements from in‐vivo human brain, as well as simulated diffusion data, using both traditional model fitting and supervised ML. For supervised ML, we train several artificial neural networks, as well as random forest regressors, on different distributions of ground truth parameters. We compare the accuracy and precision of parameter estimates obtained from the different estimation approaches using synthetic test data.
Results
When the distribution of parameter combinations in the training set matches those observed in healthy human data sets, we observe high precision, but inaccurate estimates for atypical parameter combinations. In contrast, when training data is sampled uniformly from the entire plausible parameter space, estimates tend to be more accurate for atypical parameter combinations but may have lower precision for typical parameter combinations.
Conclusion
This work highlights that estimation of model parameters using supervised ML depends strongly on the training‐set distribution. We show that high precision obtained using ML may mask strong bias, and visual assessment of the parameter maps is not sufficient for evaluating the quality of the estimates.
INTRODUCTION Many institutions have open access (OA) policies that require faculty members to deposit their articles in an institutional repository (IR). A clear motivation is that a policy will result in increased self-archiving. The purpose of this longitudinal study is to compare the impact of a campus-wide OA policy and mediated solicitation of author manuscripts, using quantitative analysis to determine the rate of article deposits over time. METHODS Metadata for faculty articles published by authors at Oregon State University between 2011 and 2014 was produced by integrating citation metadata from a bibliographic database and the IR. Author names, affiliations, and other metadata were parsed and matched to compare rates of deposit for three separate time periods relating to different OA promotional strategies. RESULTS Direct solicitation of author manuscripts is more successful in facilitating OA than an OA policy—by number of articles deposited as well as the number of unique authors participating. Author affiliation and research areas also have an impact on faculty participation in OA. DISCUSSION Outreach to colleges and departments has had a positive effect on rate of deposit for those communities of scholars. Additionally, disciplinary practice may have more influence on its members’ participation in OA. CONCLUSION Until more federal policies require open access to articles funded by grants, or institutional policies are in place that require article deposit for promotion and tenure, policies will only be as effective as the library mediated processes that are put in place to identify and solicit articles from faculty.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.