{"id":1164,"date":"2017-05-15T13:37:48","date_gmt":"2017-05-15T18:37:48","guid":{"rendered":"https:\/\/my.vanderbilt.edu\/masi\/?p=1164"},"modified":"2017-05-15T13:37:48","modified_gmt":"2017-05-15T18:37:48","slug":"accurate-age-estimation-in-a-pediatric-population-using-deep-learning-on-t1-%e2%80%90weighted-mri-structural-features","status":"publish","type":"post","link":"https:\/\/my.vanderbilt.edu\/masi\/2017\/05\/accurate-age-estimation-in-a-pediatric-population-using-deep-learning-on-t1-%e2%80%90weighted-mri-structural-features\/","title":{"rendered":"Accurate Age Estimation in a Pediatric Population Using Deep Learning on T1\u2010weighted MRI Structural Features"},"content":{"rendered":"

Citation: Bermudez, C. et.al. Accurate Age Estimation in a Pediatric Population Using Deep Learning on T1\u2010weighted MRI\u00a0 Structural Features. Frontiers i<\/span>n Biomedical Imaging Science VI. <\/span><\/em><\/strong>May 2017. Abstract.
\n<\/span><\/p>\n

Abstrract<\/p>\n

It is well known that there are structural changes that occur in the brain with age. However, there are insufficient imaging biomarkers that reliably describe structural changes in the brain, particularly in young individuals. Age estimation from structural MRI is a technique that has been shown to serve as a biomarker for aging in adults as well as diseases such as Alzheimer\u2019s Disease. This is an interesting in pediatrics because it can help generate normative curves to track neurodevelopment as well as serving as a biomarker for diagnosis and progression of neurodevelopmental diseases such as autism, dyslexia, or ADHD. Recent advances in data sharing has allowed access to large datasets of healthy, normal controls in a pediatric population in order to perform large cross-\u00adsectional analysis. Past studies in age prediction have used raw voxel intensity data of T1-\u00adweighted images to predict age with accuracy within 5-\u00ad10 years. However, these studies ignore features that have been optimized for many years in the field of image processing. In this study, we show that a deep neural networks approach on structural volumetric features derived from T1-\u00adweighted MRI outperforms random forest or classic neural networks in the task of age prediction. We apply this technique to a population of 3348 subjects ages 4-\u00ad26 years old from multiple sites and generate a quantitative biomarker called Brain Age Gap (BAG). Deep neural networks had a BAG of 2.87 years compared to 2.77 years with a classic neural network and 2.94 years with a random forest. However, we show improved predictive performance in all models after using ensemble methods, with an MAE of 2.38 years with a deep neural network ensemble, 2.51 years with a classic neural network ensemble, and 2.93 years with a random forest ensemble. This study suggests that age prediction can be accurately achieved with unimodal imaging in a young population using engineered features instead of raw images. Additionally, it provides a reliable biomarker for neurodevelopment and disease detection that can be easily translatable to the bedside.<\/p>\n","protected":false},"excerpt":{"rendered":"

Citation: Bermudez, C. et.al. Accurate Age Estimation in a Pediatric Population Using Deep Learning on T1\u2010weighted MRI\u00a0 Structural Features. Frontiers in Biomedical Imaging Science VI. May 2017. Abstract. Abstrract It is well known that there are structural changes that occur in the brain with age. However, there are insufficient imaging biomarkers that reliably describe structural…<\/p>\n","protected":false},"author":6301,"featured_media":1167,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[27,116,3,23,46,4],"tags":[],"class_list":["post-1164","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-big-data","category-image-processing","category-image-segmentation","category-machine-learning","category-multi-atlas-segmentation","category-neuroimaging"],"_links":{"self":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/1164","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\/6301"}],"replies":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/comments?post=1164"}],"version-history":[{"count":3,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/1164\/revisions"}],"predecessor-version":[{"id":1168,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/posts\/1164\/revisions\/1168"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media\/1167"}],"wp:attachment":[{"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/media?parent=1164"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/categories?post=1164"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/my.vanderbilt.edu\/masi\/wp-json\/wp\/v2\/tags?post=1164"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}