Abstract<\/h2>\n
https:\/\/pubmed.ncbi.nlm.nih.gov\/32775501\/<\/p>\n
Purpose: Deep learning methods have become essential tools for quantitative interpretation of Yucheng Tang, Riqiang Gao, Yunqiang Chen, Dashan Gao, Michael R. Savona, Richard G. Abramson, Shunxing Bao, Yuankai Huo and Bennett A. Landman, “Learning from Dispersed Manual Annotations with an Optimized Data Weighting Policy\u201d, Journal of Medical Imaging, 2020. Full Text: Abstract https:\/\/pubmed.ncbi.nlm.nih.gov\/32775501\/ Purpose: Deep learning methods have become essential tools for quantitative interpretation of medical…<\/p>\n","protected":false},"author":7582,"featured_media":2679,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5,27,60,116,23],"tags":[26,79,137],"class_list":["post-2678","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-abdomen-imaging","category-big-data","category-computed-tomography","category-image-processing","category-machine-learning","tag-abdomen","tag-computed-tomography","tag-deep-learning"],"_links":{"self":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/2678","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/users\/7582"}],"replies":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/comments?post=2678"}],"version-history":[{"count":1,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/2678\/revisions"}],"predecessor-version":[{"id":2680,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/2678\/revisions\/2680"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media\/2679"}],"wp:attachment":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media?parent=2678"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/categories?post=2678"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/tags?post=2678"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\nmedical imaging data, but training these approaches is highly sensitive to biases and class imbalance
\nin the available data. There is an opportunity to increase the available training data by
\ncombining across different data sources (e.g., distinct public projects); however, data collected
\nunder different scopes tend to have differences in class balance, label availability, and subject
\ndemographics. Recent work has shown that importance sampling can be used to guide training
\nselection. To date, these approaches have not considered imbalanced data sources with distinct
\nlabeling protocols.
\nApproach: We propose a sampling policy, known as adaptive stochastic policy (ASP), inspired
\nby reinforcement learning to adapt training based on subject, data source, and dynamic use
\ncriteria. We apply ASP in the context of multiorgan abdominal computed tomography segmentation.
\nTraining was performed with cross validation on 840 subjects from 10 data sources.
\nExternal validation was performed with 20 subjects from 1 data source.
\nResults: Four alternative strategies were evaluated with the state-of-the-art baseline as upper
\nconfident bound (UCB). ASP achieves average Dice of 0.8261 compared to 0.8135 UCB
\n(p < 0.01, paired t-test) across fivefold cross validation. On withheld testing datasets, the proposed
\nASP achieved 0.8265 mean Dice versus 0.8077 UCB (p < 0.01, paired t-test).
\nConclusions: ASP provides a flexible reweighting technique for training deep learning models.
\nWe conclude that the proposed method adapts the sample importance, which leverages the performance
\non a challenging multisite, multiorgan, and multisize segmentation task.<\/p>\n![\"learning](\"https:\/\/my.vanderbilt.edu\/masi\/wp-content\/uploads\/sites\/2304\n2661\/2020\/12\/dispersed-650x295.png\")