Overview Deep Learning trains neural networks on large datasets, including images, to generate algorithms that predict endpoints. Total Body DXA scans contain digital information about body composition (fat, bone, lean tissue) and its distribution and body shape. Each of these characteristics (bone mass, fat mass, and lean mass) has been shown to be individually predictive of clinical outcomes associated with aging.
In this project, we attempt to use deep learning methods on total body DXA scans to extract more information than was previously done, and thus, providing more accurate predictions of clinical outcomes, including cardiovascular disease (CVD), CVD death, overall mortality, cancer, cancer death, hip fracture, physical disability, incident insulin-resistant diabetes, and severity of insulin resistance.
We will use a novel new approach called self-supervised learning to extract features from the DXA whole body images collected in the Health, Aging and Body Composition (Health ABC) Study . This dataset includes more than 3,000 participants, along with their follow-up examinations in years 3, 6, and 10. This learning process does not use the aging outcomes, just the “unlabeled” images themselves. It will create a penultimate set of features found in the DXA images that can be used in the neural network or outside it to predict the outcomes of interests.
Objective/Aims The focus of this
project is to apply deep learning methods to baseline and follow-up total Body
DXA images to predict outcomes associated with age-related changes in body
composition.
Our specific aims
are:
Generate predictive algorithms for CVD, CVD death, overall mortality, major disease-free mortality, cancer, cancer death, hip fracture, physical disability, incident insulin-resistant diabetes, and hemoglobin A1c, and fasting insulin and glucose levels by applying deep learning methods to baseline and follow-up Total Body DXA images. Generate predictive algorithms for physical performance including gait speed and long-distance corridor walk speed and levels of inflammation (eg. IL-6 concentration). Explore features of TBSXA images that account for the predictive accuracy of the algorithms by using saliency mapping. Research Team Steve Cummings, MD, FAPC Principal Investigator SF Coordinating Center
John Shepherd, PhD Principal Investigator UH Cancer Center
Peter Sadowski , PhDCo-Investigator UH Manoa Information and Computer Sciences
Warren Browner, MD, MPH CEO California Pacific Medical Center
Eleanor Simonsick, PhD Epidemiologist National Institute on Aging – NIH
Yannik Glaser Graduate Student UH Manoa Information and Computer Sciences
Lily Liu Statistician California Pacific Medical Center
Funding Source(s)
Publications & Presentations 2318789 TBDXAI items 1 nlm 10 date desc 1 1 1 title https://shepherdresearchlab.org/wp-content/plugins/zotpress/ %7B%22status%22%3A%22success%22%2C%22instance%22%3A%22zotpress-905669cc445b1433f4df2f416bf4c3b6%22%2C%22meta%22%3A%7B%22request_last%22%3A0%2C%22request_next%22%3A0%2C%22used_cache%22%3Atrue%7D%2C%22data%22%3A%5B%7B%22key%22%3A%22XBGSZL6G%22%2C%22library%22%3A%7B%22id%22%3A2318789%7D%2C%22meta%22%3A%7B%22creatorSummary%22%3A%22Glaser%20et%20al.%22%2C%22parsedDate%22%3A%222022-08-16%22%2C%22numChildren%22%3A1%7D%2C%22bib%22%3A%22%3Cdiv%20class%3D%5C%22csl-bib-body%5C%22%20style%3D%5C%22line-height%3A%201.35%3B%20%5C%22%3E%5Cn%20%3Cdiv%20class%3D%5C%22csl-entry%5C%22%20style%3D%5C%22clear%3A%20left%3B%20%5C%22%3E%5Cn%20%3Cdiv%20class%3D%5C%22csl-left-margin%5C%22%20style%3D%5C%22float%3A%20left%3B%20padding-right%3A%200.5em%3B%20text-align%3A%20right%3B%20width%3A%201em%3B%5C%22%3E1.%3C%5C%2Fdiv%3E%3Cdiv%20class%3D%5C%22csl-right-inline%5C%22%20style%3D%5C%22margin%3A%200%20.4em%200%201.5em%3B%5C%22%3EGlaser%20Y%2C%20Shepherd%20J%2C%20Leong%20L%2C%20Wolfgruber%20T%2C%20Lui%20LY%2C%20Sadowski%20P%2C%20et%20al.%20%3Ca%20target%3D%27_blank%27%20href%3D%27https%3A%5C%2F%5C%2Fwww.nature.com%5C%2Farticles%5C%2Fs43856-022-00166-9%23citeas%27%3EDeep%20learning%20predicts%20all-cause%20mortality%20from%20longitudinal%20total-body%20DXA%20imaging%3C%5C%2Fa%3E.%20Commun%20Med%20%5BInternet%5D.%202022%20Aug%2016%3B2.%20%3Ca%20title%3D%27Cite%20in%20RIS%20Format%27%20class%3D%27zp-CiteRIS%27%20href%3D%27https%3A%5C%2F%5C%2Fshepherdresearchlab.org%5C%2Fwp-content%5C%2Fplugins%5C%2Fzotpress%5C%2Flib%5C%2Frequest%5C%2Frequest.cite.php%3Fapi_user_id%3D2318789%26amp%3Bitem_key%3DXBGSZL6G%27%3ECite%3C%5C%2Fa%3E%20%20%3Ca%20title%3D%27Download%27%20class%3D%27zp-DownloadURL%27%20href%3D%27https%3A%5C%2F%5C%2Fshepherdresearchlab.org%5C%2Fwp-content%5C%2Fplugins%5C%2Fzotpress%5C%2Flib%5C%2Frequest%5C%2Frequest.dl.php%3Fapi_user_id%3D2318789%26amp%3Bdlkey%3DSWULZ2KQ%26amp%3Bcontent_type%3Dapplication%5C%2Fpdf%27%3EDownload%3C%5C%2Fa%3E%3C%5C%2Fdiv%3E%5Cn%20%3C%5C%2Fdiv%3E%5Cn%3C%5C%2Fdiv%3E%22%2C%22data%22%3A%7B%22itemType%22%3A%22journalArticle%22%2C%22title%22%3A%22Deep%20learning%20predicts%20all-cause%20mortality%20from%20longitudinal%20total-body%20DXA%20imaging%22%2C%22creators%22%3A%5B%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Yannik%22%2C%22lastName%22%3A%22Glaser%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22John%22%2C%22lastName%22%3A%22Shepherd%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Lambert%22%2C%22lastName%22%3A%22Leong%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Thomas%22%2C%22lastName%22%3A%22Wolfgruber%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Li-Yung%22%2C%22lastName%22%3A%22Lui%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Peter%22%2C%22lastName%22%3A%22Sadowski%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Steven%20R%22%2C%22lastName%22%3A%22Cummings%22%7D%5D%2C%22abstractNote%22%3A%22Mortality%20research%20has%20identified%20biomarkers%20predictive%20of%20all-cause%20mortality%20risk.%20Most%20of%20these%20markers%2C%20such%20as%5Cnpatient%20body%20mass%20index%2C%20are%20predictive%20cross-sectionally%20with%20a%20single%20datapoint%2C%20and%20for%20others%20the%20longitudinal%20change%5Cnhas%20been%20shown%20to%20be%20predictive%2C%20for%20instance%20greater-than-average%20muscle%20and%20weight%20loss%20in%20older%20adults.%20And%20while%5Cnsometimes%20markers%20are%20derived%20from%20imaging%20modalities%20such%20as%20DXA%2C%20the%20full%20scans%20are%20rarely%20used.%5CnThis%20study%20builds%20on%20that%20knowledge%20and%20tests%20two%20hypotheses%20to%20improve%20all-cause%20mortality%20prediction%3A%201%29%20risk%20features%5Cnderived%20from%20raw%20total-body%20DXA%20imaging%20data%20using%20deep%20learning%20are%20predictive%20of%20all-cause%20mortality%20with%20and%20without%5Cnclinical%20risk%20factors%2C%20and%202%29%20sequential%20total-body%20DXA%20scans%20and%20recurrent%20neural%20network%20models%20outperform%5Cncomparable%20models%20using%20only%20one%20observation%20with%20and%20without%20clinical%20risk%20factors.%20Multiple%20deep%20neural%20network%5Cnarchitectures%20were%20designed%20to%20test%20theses%20hypotheses.%20The%20results%20demonstrate%20that%20longitudinal%20total-body%20DXA%20scans%5Cnwere%20predictive%20of%20all-cause%20mortality%20and%20improved%20performance%20of%20traditional%20mortality%20prediction%20models.%20The%20models%5Cnwere%20trained%20and%20evaluated%20on%20data%20from%20the%2016-year-long%20Health%2C%20Aging%2C%20and%20Body%20Composition%20Study%20including%20over%5Cn15%2C000%20scans%20from%20over%203%2C000%20older%20multi-race%20male%20and%20female%20adults.%20On%20a%20held-out%20test%20set%2C%20the%20strongest%20model%5Cnachieves%20an%20area%20under%20the%20receiver%20operator%20characteristic%20curve%20of%200.79.%5CnThis%20study%20further%20used%20explainable%20AI%20techniques%20to%20interpret%20the%20predictions%20and%20evaluate%20the%20contribution%20of%20different%5Cninputs.%20This%20approach%20sheds%20light%20on%20what%20constitutes%20healthy%20aging%20and%20simultaneously%20paves%20the%20way%20for%20new%20clinical%5Cnintervention%20strategies%20based%20on%20highly%20personalized%20healthcare%20insights.%22%2C%22date%22%3A%2216%20August%202022%22%2C%22language%22%3A%22%22%2C%22DOI%22%3A%2210.1038%5C%2Fs43856-022-00166-9%22%2C%22ISSN%22%3A%22%22%2C%22url%22%3A%22https%3A%5C%2F%5C%2Fwww.nature.com%5C%2Farticles%5C%2Fs43856-022-00166-9%23citeas%22%2C%22collections%22%3A%5B%22SJAER8EK%22%2C%228UXNM7I7%22%5D%2C%22dateModified%22%3A%222022-08-24T02%3A56%3A13Z%22%7D%7D%2C%7B%22key%22%3A%22BQ8EJIV6%22%2C%22library%22%3A%7B%22id%22%3A2318789%7D%2C%22meta%22%3A%7B%22creatorSummary%22%3A%22Glaser%20et%20al.%22%2C%22parsedDate%22%3A%222020-09-11%22%2C%22numChildren%22%3A1%7D%2C%22bib%22%3A%22%3Cdiv%20class%3D%5C%22csl-bib-body%5C%22%20style%3D%5C%22line-height%3A%201.35%3B%20%5C%22%3E%5Cn%20%20%3Cdiv%20class%3D%5C%22csl-entry%5C%22%20style%3D%5C%22clear%3A%20left%3B%20%5C%22%3E%5Cn%20%20%20%20%3Cdiv%20class%3D%5C%22csl-left-margin%5C%22%20style%3D%5C%22float%3A%20left%3B%20padding-right%3A%200.5em%3B%20text-align%3A%20right%3B%20width%3A%201em%3B%5C%22%3E1.%3C%5C%2Fdiv%3E%3Cdiv%20class%3D%5C%22csl-right-inline%5C%22%20style%3D%5C%22margin%3A%200%20.4em%200%201.5em%3B%5C%22%3EGlaser%20Y%2C%20Sadowski%20P%2C%20Wolfgruber%20T%2C%20Lui%20LY%2C%20Cummings%20S%2C%20Shepherd%20J.%20Hip%20Fracture%20Risk%20Modelling%20Using%20DXA%20and%20Artificial%20Intelligence.%20Poster%20presented%20at%3A%20American%20Society%20for%20Bone%20Mineral%20and%20Research%20Annual%20Meeting%3B%202020%20Sep%2011%3B%20Virtual.%20%3Ca%20title%3D%27Cite%20in%20RIS%20Format%27%20class%3D%27zp-CiteRIS%27%20href%3D%27https%3A%5C%2F%5C%2Fshepherdresearchlab.org%5C%2Fwp-content%5C%2Fplugins%5C%2Fzotpress%5C%2Flib%5C%2Frequest%5C%2Frequest.cite.php%3Fapi_user_id%3D2318789%26amp%3Bitem_key%3DBQ8EJIV6%27%3ECite%3C%5C%2Fa%3E%20%20%3Ca%20title%3D%27Download%27%20class%3D%27zp-DownloadURL%27%20href%3D%27https%3A%5C%2F%5C%2Fshepherdresearchlab.org%5C%2Fwp-content%5C%2Fplugins%5C%2Fzotpress%5C%2Flib%5C%2Frequest%5C%2Frequest.dl.php%3Fapi_user_id%3D2318789%26amp%3Bdlkey%3DYU5FVCPN%26amp%3Bcontent_type%3Dapplication%5C%2Fpdf%27%3EDownload%3C%5C%2Fa%3E%3C%5C%2Fdiv%3E%5Cn%20%20%20%3C%5C%2Fdiv%3E%5Cn%3C%5C%2Fdiv%3E%22%2C%22data%22%3A%7B%22itemType%22%3A%22presentation%22%2C%22title%22%3A%22Hip%20Fracture%20Risk%20Modelling%20Using%20DXA%20and%20Artificial%20Intelligence%22%2C%22creators%22%3A%5B%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Yannik%22%2C%22lastName%22%3A%22Glaser%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Peter%22%2C%22lastName%22%3A%22Sadowski%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Thomas%22%2C%22lastName%22%3A%22Wolfgruber%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Li-Yung%22%2C%22lastName%22%3A%22Lui%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Steve%22%2C%22lastName%22%3A%22Cummings%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22John%22%2C%22lastName%22%3A%22Shepherd%22%7D%5D%2C%22abstractNote%22%3A%22%22%2C%22date%22%3A%22September%2011-15%2C%202020%22%2C%22url%22%3A%22%22%2C%22language%22%3A%22%22%2C%22collections%22%3A%5B%22US56U5WJ%22%2C%22IDI3J9GR%22%5D%2C%22dateModified%22%3A%222022-07-09T02%3A30%3A54Z%22%7D%7D%5D%7D 1.
Glaser Y, Sadowski P, Wolfgruber T, Lui LY, Cummings S, Shepherd J. Hip Fracture Risk Modelling Using DXA and Artificial Intelligence. Poster presented at: American Society for Bone Mineral and Research Annual Meeting; 2020 Sep 11; Virtual.
Cite Download