I’m a PhD student in the Computer Science (ICS) department at the University of Hawai’i. I have Bachelors degrees in Statistics and Computer Science, and am passionate about the application of data science to healthcare. I currently work as a Research Assistant in the UH Machine Learning Lab, working in applied ML. My work with the Shepherd Research Lab focuses on applying deep learning to breast ultrasound imaging.
Presentations & Publications
2318789 Bunnell items 1 nlm 7 year desc 1 1 title Bunnell A https://shepherdresearchlab.org/wp-content/plugins/zotpress/
%7B%22status%22%3A%22success%22%2C%22instance%22%3A%22zotpress-6d0206ad263ab276a86d354f2db7320a%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%22ZRUYIMT4%22%2C%22library%22%3A%7B%22id%22%3A2318789%7D%2C%22meta%22%3A%7B%22creatorSummary%22%3A%22Bunnell%20et%20al.%22%2C%22parsedDate%22%3A%222023-03-01%22%2C%22numChildren%22%3A0%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%3E%3Cstrong%3EBunnell%20A%3C%5C%2Fstrong%3E%2C%20Valdez%20D%2C%20Wolfgruber%20T%2C%20Altamirano%20A%2C%20Hernandez%20B%2C%20Sadowski%20P%2C%20et%20al.%20%3Ca%20href%3D%27https%3A%5C%2F%5C%2Fdoi.org%5C%2F10.1158%5C%2F1538-7445.SABCS22-P3-04-05%27%3EAbstract%20P3-04-05%3A%20Artificial%20Intelligence%20Detects%2C%20Classifies%2C%20and%20Describes%20Lesions%20in%20Clinical%20Breast%20Ultrasound%20Images%3C%5C%2Fa%3E%20%5BInternet%5D.%20San%20Antonio%20Breast%20Cancer%20Symposium%202022%3B%202023%20Mar%201%20%5Bcited%202023%20Mar%2015%5D.%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%3DZRUYIMT4%27%3ECite%3C%5C%2Fa%3E%20%3C%5C%2Fdiv%3E%5Cn%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%22Abstract%20P3-04-05%3A%20Artificial%20Intelligence%20Detects%2C%20Classifies%2C%20and%20Describes%20Lesions%20in%20Clinical%20Breast%20Ultrasound%20Images%22%2C%22creators%22%3A%5B%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Arianna%22%2C%22lastName%22%3A%22Bunnell%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Dustin%22%2C%22lastName%22%3A%22Valdez%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%22Aleen%22%2C%22lastName%22%3A%22Altamirano%22%7D%2C%7B%22creatorType%22%3A%22presenter%22%2C%22firstName%22%3A%22Brenda%22%2C%22lastName%22%3A%22Hernandez%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%22John%22%2C%22lastName%22%3A%22Shepherd%22%7D%5D%2C%22abstractNote%22%3A%22Purpose%20Many%20low-middle%20income%20countries%20%28LMIC%29%20suffer%20from%20chronic%20shortages%20of%20resources%20that%20inhibit%20the%20implementation%20of%20effective%20breast%20cancer%20screening%20programs.%20Advanced%20breast%20cancer%20rates%20in%20the%20U.S.%20Affiliated%20Pacific%20Islands%20substantially%20exceed%20that%20of%20the%20United%20States.%20We%20propose%20the%20use%20of%20portable%20breast%20ultrasound%20coupled%20with%20artificial%20intelligence%20%28AI%29%20algorithms%20to%20assist%20non-radiologist%20field%20personnel%20in%20real-time%20field%20lesion%20detection%2C%20classification%2C%20and%20biopsy%2C%20as%20well%20as%20determination%20of%20breast%20density%20for%20risk%20assessment.%20In%20this%20study%2C%20we%20examine%20the%20ability%20of%20an%20AI%20algorithm%20to%20detect%20and%20describe%20breast%20cancer%20lesions%20in%20clinically-acquired%20breast%20ultrasound%20images%20in%2040%2C000%20women%20participating%20in%20a%20Hawaii%20screening%20program.%20Materials%20and%20Methods%20The%20Hawaii%20and%20Pacific%20Islands%20Mammography%20Registry%20%28HIPIMR%29%20collects%20breast%20health%20questionnaires%20and%20breast%20imaging%20%28mammography%2C%20ultrasound%2C%20and%20MRI%29%20from%20participating%20centers%20in%20Hawaii%20and%20the%20Pacific%20and%20links%20this%20information%20to%20the%20Hawaii%20Tumor%20Registry%20for%20cancer%20findings.%20From%20the%20women%20with%20either%20screening%20or%20diagnostic%20B-mode%20breast%20ultrasound%20exams%2C%20we%20selected%203%20negative%20cases%20%28no%20cancer%29%20for%20every%20positive%20case%20matched%20by%20age%2C%20excluding%20Doppler%20and%20elastography%20images.%20The%20blinded%20images%20were%20read%20by%20the%20study%20radiologist%20to%20delineate%20all%20lesions%20and%20describe%20in%20terms%20of%20the%20BI-RADS%20lexicon.%20These%20images%20were%20split%20by%20woman%20into%20training%20%2870%25%29%2C%20validation%20and%20hyperparameter%20selection%20%2820%25%29%20and%20testing%20%2820%25%29%20subsets.%20An%20AI%20model%20was%20fine-tuned%20for%20lesion%20and%20BI-RADS%20category%20classification%20from%20a%20Detectron2%20framework%20%5B1%5D%20pre-trained%20on%20the%20COCO%20Instance%20Segmentation%20Dataset%20%5B2%5D.%20Model%20performance%20was%20evaluated%20by%20computation%20of%20precision%20and%20sensitivity%20percentages%2C%20as%20well%20as%20Area%20under%20the%20Receiver%20Operator%20Curve%20%28AUROC%29.%20Detections%20were%20considered%20positive%20if%20they%20overlapped%20a%20ground%20truth%20lesion%20delineation%20by%20at%20least%2050%25%20%28Intersection%20over%20Union%20%3D%200.5%29%2C%20and%20a%20maximum%20of%204%20detections%20were%20generated%20for%20each%20image.%20Timing%20experiments%20were%20run%20on%20a%20GPU-enabled%20%28Nvidia%20Tesla%20V100%29%20machine%20on%20unbatched%20images.%20Results%20Over%20the%2010-year%20observation%20period%2C%20we%20identified%205%2C214%20women%20with%20US%20images%20meeting%20our%20criterion.%20Of%20these%2C%20111%20were%20diagnosed%20with%20malignant%20breast%20cancer%20and%20333%20were%20selected%20as%20non-cases%20for%20a%20total%20of%20444%20women.%20These%20444%20women%20had%20a%20total%20of%204%2C623%20ultrasound%20images%20with%202%2C028%20benign%20and%201%2C431%20malignant%20lesions%20identified%20by%20the%20study%20radiologist.%20For%20cancerous%20lesions%2C%20the%20AI%20algorithm%20had%208%25%20precision%20at%20a%20sensitivity%20of%2090%25%20on%20the%20testing%20set.%20For%20benign%20lesions%2C%20a%20sensitivity%20of%2090%25%20resulted%20in%205%25%20precision%20on%20the%20testing%20set.%20The%20AUROC%20for%20bounding%20box%20detections%20of%20cancerous%20lesions%20was%200.90.%20The%20AUROC%20for%20bounding%20box%20detections%20of%20benign%20lesions%20was%200.87.%20The%20model%20made%20predictions%20at%20a%20rate%20of%2025%20frames%5C%2Fsecond%20time%20%2838.7%20milliseconds%20per%20image%29.%20Conclusion%20Detection%2C%20segmentation%2C%20and%20cancer%20classification%20of%20breast%20lesions%20are%20possible%20in%20clinically-acquired%20ultrasound%20images%20using%20AI.%20Based%20on%20our%20timing%20experiments%2C%20the%20model%20is%20capable%20of%20detecting%20and%20classifying%20lesions%20in%20real-time%20during%20ultrasound%20capture.%20Model%20performance%20is%20expected%20to%20improve%20as%20more%20data%20becomes%20available%20for%20training.%20Future%20work%20would%20involve%20further%20fine-tuning%20of%20the%20model%20on%20portable%20breast%20ultrasound%20images%20and%20increasing%20model%20evaluation%20speed%20in%20order%20to%20assess%20utility%20in%20low-resource%20populations%20%5B1%5D%20Wu%20Y%2C%20Kirillov%20A%2C%20Massa%20F%2C%20Lo%20W-Y%2C%20Girshick%20R.%20Detectron2.%20https%3A%5C%2F%5C%2Fgithub.com%5C%2Ffacebookresearch%5C%2Fdetectron2.%20%5B2%5D%20Lin%20T-Y%2C%20Maire%20M%2C%20Belongie%20S%2C%20et%20al.%20Microsoft%20COCO%3A%20Common%20Objects%20in%20Context.%20Computer%20Vision%20%5Cu2013%20ECCV%202014.%20Springer%20International%20Publishing%3B%202014%3A740-755.Citation%20Format%3A%20Arianna%20Bunnell%2C%20Dustin%20Valdez%2C%20Thomas%20Wolfgruber%2C%20Aleen%20Altamirano%2C%20Brenda%20Hernandez%2C%20Peter%20Sadowski%2C%20John%20Shepherd.%20Artificial%20Intelligence%20Detects%2C%20Classifies%2C%20and%20Describes%20Lesions%20in%20Clinical%20Breast%20Ultrasound%20Images%20%5Babstract%5D.%20In%3A%20Proceedings%20of%20the%202022%20San%20Antonio%20Breast%20Cancer%20Symposium%3B%202022%20Dec%206-10%3B%20San%20Antonio%2C%20TX.%20Philadelphia%20%28PA%29%3A%20AACR%3B%20Cancer%20Res%202023%3B83%285%20Suppl%29%3AAbstract%20nr%20P3-04-05.%22%2C%22date%22%3A%222023-03-01%22%2C%22url%22%3A%22https%3A%5C%2F%5C%2Fdoi.org%5C%2F10.1158%5C%2F1538-7445.SABCS22-P3-04-05%22%2C%22language%22%3A%22%22%2C%22collections%22%3A%5B%22IDI3J9GR%22%2C%22A2J6KQIV%22%5D%2C%22dateModified%22%3A%222023-08-02T00%3A18%3A52Z%22%7D%7D%2C%7B%22key%22%3A%22X23IRV7C%22%2C%22library%22%3A%7B%22id%22%3A2318789%7D%2C%22meta%22%3A%7B%22creatorSummary%22%3A%22Valdez%20et%20al.%22%2C%22parsedDate%22%3A%222023-03-01%22%2C%22numChildren%22%3A0%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%3EValdez%20D%2C%20%3Cstrong%3EBunnell%20A%3C%5C%2Fstrong%3E%2C%20Wolfgruber%20T%2C%20Altamirano%20A%2C%20Quon%20B%2C%20Maskarinec%20G%2C%20et%20al.%20%3Ca%20href%3D%27https%3A%5C%2F%5C%2Fdoi.org%5C%2F10.1158%5C%2F1538-7445.SABCS22-P3-03-02%27%3EAbstract%20P3-03-02%3A%20Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%3C%5C%2Fa%3E.%20Cancer%20Research%20%5BInternet%5D.%202023%20Mar%201%20%5Bcited%202023%20Jun%2021%5D%3B83%285_Supplement%29%3AP3-03%5Cu201302.%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%3DX23IRV7C%27%3ECite%3C%5C%2Fa%3E%20%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%22Abstract%20P3-03-02%3A%20Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%22%2C%22creators%22%3A%5B%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Dustin%22%2C%22lastName%22%3A%22Valdez%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Arianna%22%2C%22lastName%22%3A%22Bunnell%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%22Aleen%22%2C%22lastName%22%3A%22Altamirano%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Brandon%22%2C%22lastName%22%3A%22Quon%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Gertraud%22%2C%22lastName%22%3A%22Maskarinec%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%22John%22%2C%22lastName%22%3A%22Shepherd%22%7D%5D%2C%22abstractNote%22%3A%22Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%20Dustin%20Valdez12%2C%20Arianna%20Bunnell2%2C%20Thomas%20Wolfgruber1%2C%20Aleen%20V.%20Altamirano3%2C%20Brandon%20Quon1%2C%20Gertraud%20Maskarinec1%2C%20Peter%20Sadowski2%2C%20John%20A.%20Shepherd1%201%20University%20of%20Hawaii%20Cancer%20Center%2C%20Honolulu%2C%20HI%202%20University%20of%20Hawaii%20at%20Manoa%2C%20Honolulu%2C%20HI%203%20Instituto%20Radiodiagn%5Cu00f3stico%2C%20Managua%2C%20Nicaragua%20Background%3A%20Breast%20cancer%20is%20the%20second%20leading%20cause%20of%20cancer-related%20death%20among%20women%20in%20Hawaii%20and%20the%20Pacific.%20However%2C%20while%20there%20are%20programs%20like%20the%20Breast%20and%20Cervical%20Cancer%20Early%20Detection%20Program%20%28BCCEDP%29%20implemented%20throughout%20the%20Pacific%2C%20the%20lack%20of%20access%20to%20mammography%20screening%20and%20low%20screening%20participation%20rates%20contributes%20to%20very%20high%20advanced%20breast%20cancer%20rates%20in%20most%20cases%20over%2050%25.%20Portable%20breast%20ultrasound%20is%20a%20promising%20screening%20technology%20for%20low%20resource%20areas.%20However%2C%20without%20mammography%2C%20mammographic%20density%20is%20not%20available%20for%20risk%20modeling%20to%20determine%20who%20should%20participate%20in%20screening%20programs%20or%20at%20what%20frequency.%20In%20this%20study%2C%20we%20ask%20if%20breast%20ultrasound%20%28US%29%20images%20can%20be%20used%20to%20derive%20an%20equivalent%20mammographic%20density%20for%20risk%20modeling.%20We%20utilized%20artificial%20intelligence%20to%20derive%20breast%20density%20from%20diagnostic%20ultrasound%20images%20and%20compared%20to%20BI-RADS%20mammographic%20density%20in%20an%20established%20breast%20cancer%20risk%20model1.%20Methods%3A%20We%20selected%20women%20with%20negative%20screening%20visit%20who%20either%20later%20developed%20cancer%20%28positives%29%20or%20did%20not%20%28negatives%29%20over%20a%2010-year%20period.%20Temporally-matched%20negative%20mammographic%20and%20ultrasound%20images%2C%20cancer%20outcome%20status%20and%20cancer%20risk%20information%20were%20sourced%20from%20the%20Hawaii%20and%20Pacific%20Islands%20Mammography%20Registry.%20US%20images%20had%20to%20have%20occurred%20within%20a%20year%20of%20the%20mammogram.%20BI-RADS%20mammographic%20density%20was%20derived%20using%20an%20existing%20deep%20neural%20network%20model2.%20Mammographic%20density%20was%20estimated%20from%20US%20images%20by%20training%20a%20deep-learning%20convolutional%20neural%20network%20model.%20A%20hold%20out%20set%20of%20images%20%28Test%20set%20of%2020%25%20of%20the%20total%29%20was%20used%20to%20compare%2010-year%20breast%20cancer%20risk%20using%20the%20Tyrer-Cuzick%20%28TC%29%20risk%20model1%20when%20calculated%20using%20breast%20density%20from%20either%20mammograms%20or%20US.%20The%20AUC%20values%2C%20confidence%20intervals%2C%20ROC%20plots%20and%20Pearson%20correlation%20were%20calculated%20and%20compared.%20Results%3A%20Over%20the%2010-year%20study%20period%2C%201337%20had%20matched%20mammograms%20and%20US%20images%20and%2065%20went%20on%20to%20develop%20breast%20cancer.%20Using%20the%20test%20set%2C%20the%20Pearson%5Cu2019s%20correlation%20between%20breast%20density%20from%20mammography%20and%20US%20was%200.31%20%28moderate%20correlation%29.%20There%20were%20no%20covariates%20found%20to%20improve%20this%20association.%20The%20AUC%20for%20TC%2010-year%20personal%20risk%20was%20higher%20when%20breast%20density%20from%20mammograms%20was%20used%200.71%20%2895%25%20CI%3D0.57-0.86%29%20versus%20US%20images%200.65%20%2895%25%20CI%3D0.53-0.76%29.%20Conclusion%3A%20Overall%20breast%20cancer%20risk%20was%20similar%20when%20breast%20density%20was%20derived%20from%20either%20mammograms%20or%20US.%20The%20performance%20of%20our%20US%20breast%20density%20model%20is%20expected%20to%20improve%20further%20when%20more%20US%20training%20data%20becomes%20available.%20Breast%20cancer%20screening%20programs%20exclusively%20using%20US%20imaging%20may%20be%20able%20to%20provide%20equivalent%20risk%20modeling%20to%20clinics%20using%20mammography.%201.Tyrer%20J%2C%20Duffy%20SW%2C%20Cuzick%20J%20%282004%29.%20A%20breast%20cancer%20prediction%20model%20incorporating%20familial%20and%20personal%20risk%20factors.%20Stat%20Med.%202004%20Apr%2015%3B23%287%29%3A1111-30.%20doi%3A%2010.1002%5C%2Fsim.1668.%20Erratum%20in%3A%20Stat%20Med.%202005%20Jan%2015%3B24%281%29%3A156.%20PMID%3A%2015057881.%202.%20Wu%2C%20N.%2C%20K.%20J.%20Geras%2C%20Y.%20Shen%2C%20J.%20Su%2C%20S.%20G.%20Kim%2C%20E.%20Kim%2C%20S.%20Wolfson%2C%20L.%20Moy%20and%20K.%20Cho%20%282018%29.%20Breast%20Density%20Classification%20with%20Deep%20Convolutional%20Neural%20Networks.%202018%20IEEE%20International%20Conference%20on%20Acoustics%2C%20Speech%20and%20Signal%20Processing%20%28ICASSP%29%2C%20IEEE.Citation%20Format%3A%20Dustin%20Valdez%2C%20Arianna%20Bunnell%2C%20Thomas%20Wolfgruber%2C%20Aleen%20Altamirano%2C%20Brandon%20Quon%2C%20Gertraud%20Maskarinec%2C%20Peter%20Sadowski%2C%20John%20Shepherd.%20Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%20%5Babstract%5D.%20In%3A%20Proceedings%20of%20the%202022%20San%20Antonio%20Breast%20Cancer%20Symposium%3B%202022%20Dec%206-10%3B%20San%20Antonio%2C%20TX.%20Philadelphia%20%28PA%29%3A%20AACR%3B%20Cancer%20Res%202023%3B83%285%20Suppl%29%3AAbstract%20nr%20P3-03-02.%22%2C%22date%22%3A%222023-03-01%22%2C%22language%22%3A%22%22%2C%22DOI%22%3A%2210.1158%5C%2F1538-7445.SABCS22-P3-03-02%22%2C%22ISSN%22%3A%220008-5472%22%2C%22url%22%3A%22https%3A%5C%2F%5C%2Fdoi.org%5C%2F10.1158%5C%2F1538-7445.SABCS22-P3-03-02%22%2C%22collections%22%3A%5B%22IDI3J9GR%22%2C%22A2J6KQIV%22%5D%2C%22dateModified%22%3A%222023-07-11T01%3A08%3A11Z%22%7D%7D%2C%7B%22key%22%3A%22WMGEY4I5%22%2C%22library%22%3A%7B%22id%22%3A2318789%7D%2C%22meta%22%3A%7B%22creatorSummary%22%3A%22Valdez%20et%20al.%22%2C%22parsedDate%22%3A%222023-03-01%22%2C%22numChildren%22%3A0%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%3EValdez%20D%2C%20%3Cstrong%3EBunnell%20A%3C%5C%2Fstrong%3E%2C%20Wolfgruber%20T%2C%20Altamirano%20A%2C%20Quon%20B%2C%20Maskarinec%20G%2C%20et%20al.%20%3Ca%20href%3D%27https%3A%5C%2F%5C%2Fdoi.org%5C%2F10.1158%5C%2F1538-7445.SABCS22-P3-03-02%27%3EAbstract%20P3-03-02%3A%20Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%3C%5C%2Fa%3E.%20Cancer%20Research%20%5BInternet%5D.%202023%20Mar%201%20%5Bcited%202023%20Mar%2015%5D%3B83%285_Supplement%29%3AP3-03%5Cu201302.%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%3DWMGEY4I5%27%3ECite%3C%5C%2Fa%3E%20%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%22Abstract%20P3-03-02%3A%20Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%22%2C%22creators%22%3A%5B%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Dustin%22%2C%22lastName%22%3A%22Valdez%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Arianna%22%2C%22lastName%22%3A%22Bunnell%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%22Aleen%22%2C%22lastName%22%3A%22Altamirano%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Brandon%22%2C%22lastName%22%3A%22Quon%22%7D%2C%7B%22creatorType%22%3A%22author%22%2C%22firstName%22%3A%22Gertraud%22%2C%22lastName%22%3A%22Maskarinec%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%22John%22%2C%22lastName%22%3A%22Shepherd%22%7D%5D%2C%22abstractNote%22%3A%22Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%20Dustin%20Valdez12%2C%20Arianna%20Bunnell2%2C%20Thomas%20Wolfgruber1%2C%20Aleen%20V.%20Altamirano3%2C%20Brandon%20Quon1%2C%20Gertraud%20Maskarinec1%2C%20Peter%20Sadowski2%2C%20John%20A.%20Shepherd1%201%20University%20of%20Hawaii%20Cancer%20Center%2C%20Honolulu%2C%20HI%202%20University%20of%20Hawaii%20at%20Manoa%2C%20Honolulu%2C%20HI%203%20Instituto%20Radiodiagn%5Cu00f3stico%2C%20Managua%2C%20Nicaragua%20Background%3A%20Breast%20cancer%20is%20the%20second%20leading%20cause%20of%20cancer-related%20death%20among%20women%20in%20Hawaii%20and%20the%20Pacific.%20However%2C%20while%20there%20are%20programs%20like%20the%20Breast%20and%20Cervical%20Cancer%20Early%20Detection%20Program%20%28BCCEDP%29%20implemented%20throughout%20the%20Pacific%2C%20the%20lack%20of%20access%20to%20mammography%20screening%20and%20low%20screening%20participation%20rates%20contributes%20to%20very%20high%20advanced%20breast%20cancer%20rates%20in%20most%20cases%20over%2050%25.%20Portable%20breast%20ultrasound%20is%20a%20promising%20screening%20technology%20for%20low%20resource%20areas.%20However%2C%20without%20mammography%2C%20mammographic%20density%20is%20not%20available%20for%20risk%20modeling%20to%20determine%20who%20should%20participate%20in%20screening%20programs%20or%20at%20what%20frequency.%20In%20this%20study%2C%20we%20ask%20if%20breast%20ultrasound%20%28US%29%20images%20can%20be%20used%20to%20derive%20an%20equivalent%20mammographic%20density%20for%20risk%20modeling.%20We%20utilized%20artificial%20intelligence%20to%20derive%20breast%20density%20from%20diagnostic%20ultrasound%20images%20and%20compared%20to%20BI-RADS%20mammographic%20density%20in%20an%20established%20breast%20cancer%20risk%20model1.%20Methods%3A%20We%20selected%20women%20with%20negative%20screening%20visit%20who%20either%20later%20developed%20cancer%20%28positives%29%20or%20did%20not%20%28negatives%29%20over%20a%2010-year%20period.%20Temporally-matched%20negative%20mammographic%20and%20ultrasound%20images%2C%20cancer%20outcome%20status%20and%20cancer%20risk%20information%20were%20sourced%20from%20the%20Hawaii%20and%20Pacific%20Islands%20Mammography%20Registry.%20US%20images%20had%20to%20have%20occurred%20within%20a%20year%20of%20the%20mammogram.%20BI-RADS%20mammographic%20density%20was%20derived%20using%20an%20existing%20deep%20neural%20network%20model2.%20Mammographic%20density%20was%20estimated%20from%20US%20images%20by%20training%20a%20deep-learning%20convolutional%20neural%20network%20model.%20A%20hold%20out%20set%20of%20images%20%28Test%20set%20of%2020%25%20of%20the%20total%29%20was%20used%20to%20compare%2010-year%20breast%20cancer%20risk%20using%20the%20Tyrer-Cuzick%20%28TC%29%20risk%20model1%20when%20calculated%20using%20breast%20density%20from%20either%20mammograms%20or%20US.%20The%20AUC%20values%2C%20confidence%20intervals%2C%20ROC%20plots%20and%20Pearson%20correlation%20were%20calculated%20and%20compared.%20Results%3A%20Over%20the%2010-year%20study%20period%2C%201337%20had%20matched%20mammograms%20and%20US%20images%20and%2065%20went%20on%20to%20develop%20breast%20cancer.%20Using%20the%20test%20set%2C%20the%20Pearson%5Cu2019s%20correlation%20between%20breast%20density%20from%20mammography%20and%20US%20was%200.31%20%28moderate%20correlation%29.%20There%20were%20no%20covariates%20found%20to%20improve%20this%20association.%20The%20AUC%20for%20TC%2010-year%20personal%20risk%20was%20higher%20when%20breast%20density%20from%20mammograms%20was%20used%200.71%20%2895%25%20CI%3D0.57-0.86%29%20versus%20US%20images%200.65%20%2895%25%20CI%3D0.53-0.76%29.%20Conclusion%3A%20Overall%20breast%20cancer%20risk%20was%20similar%20when%20breast%20density%20was%20derived%20from%20either%20mammograms%20or%20US.%20The%20performance%20of%20our%20US%20breast%20density%20model%20is%20expected%20to%20improve%20further%20when%20more%20US%20training%20data%20becomes%20available.%20Breast%20cancer%20screening%20programs%20exclusively%20using%20US%20imaging%20may%20be%20able%20to%20provide%20equivalent%20risk%20modeling%20to%20clinics%20using%20mammography.%201.Tyrer%20J%2C%20Duffy%20SW%2C%20Cuzick%20J%20%282004%29.%20A%20breast%20cancer%20prediction%20model%20incorporating%20familial%20and%20personal%20risk%20factors.%20Stat%20Med.%202004%20Apr%2015%3B23%287%29%3A1111-30.%20doi%3A%2010.1002%5C%2Fsim.1668.%20Erratum%20in%3A%20Stat%20Med.%202005%20Jan%2015%3B24%281%29%3A156.%20PMID%3A%2015057881.%202.%20Wu%2C%20N.%2C%20K.%20J.%20Geras%2C%20Y.%20Shen%2C%20J.%20Su%2C%20S.%20G.%20Kim%2C%20E.%20Kim%2C%20S.%20Wolfson%2C%20L.%20Moy%20and%20K.%20Cho%20%282018%29.%20Breast%20Density%20Classification%20with%20Deep%20Convolutional%20Neural%20Networks.%202018%20IEEE%20International%20Conference%20on%20Acoustics%2C%20Speech%20and%20Signal%20Processing%20%28ICASSP%29%2C%20IEEE.Citation%20Format%3A%20Dustin%20Valdez%2C%20Arianna%20Bunnell%2C%20Thomas%20Wolfgruber%2C%20Aleen%20Altamirano%2C%20Brandon%20Quon%2C%20Gertraud%20Maskarinec%2C%20Peter%20Sadowski%2C%20John%20Shepherd.%20Can%20artificial%20intelligence%20derived%20ultrasound%20breast%20density%20provide%20comparable%20breast%20cancer%20risk%20estimates%20to%20density%20derived%20from%20mammograms%20%5Babstract%5D.%20In%3A%20Proceedings%20of%20the%202022%20San%20Antonio%20Breast%20Cancer%20Symposium%3B%202022%20Dec%206-10%3B%20San%20Antonio%2C%20TX.%20Philadelphia%20%28PA%29%3A%20AACR%3B%20Cancer%20Res%202023%3B83%285%20Suppl%29%3AAbstract%20nr%20P3-03-02.%22%2C%22date%22%3A%222023-03-01%22%2C%22language%22%3A%22%22%2C%22DOI%22%3A%2210.1158%5C%2F1538-7445.SABCS22-P3-03-02%22%2C%22ISSN%22%3A%220008-5472%22%2C%22url%22%3A%22https%3A%5C%2F%5C%2Fdoi.org%5C%2F10.1158%5C%2F1538-7445.SABCS22-P3-03-02%22%2C%22collections%22%3A%5B%22IDI3J9GR%22%2C%22A2J6KQIV%22%5D%2C%22dateModified%22%3A%222023-03-14T01%3A18%3A26Z%22%7D%7D%5D%7D
1.
Bunnell A, Valdez D, Wolfgruber T, Altamirano A, Hernandez B, Sadowski P, et al. Abstract P3-04-05: Artificial Intelligence Detects, Classifies, and Describes Lesions in Clinical Breast Ultrasound Images [Internet]. San Antonio Breast Cancer Symposium 2022; 2023 Mar 1 [cited 2023 Mar 15]. Cite
1.
Valdez D, Bunnell A, Wolfgruber T, Altamirano A, Quon B, Maskarinec G, et al. Abstract P3-03-02: Can artificial intelligence derived ultrasound breast density provide comparable breast cancer risk estimates to density derived from mammograms. Cancer Research [Internet]. 2023 Mar 1 [cited 2023 Jun 21];83(5_Supplement):P3-03–02. Cite
1.
Valdez D, Bunnell A, Wolfgruber T, Altamirano A, Quon B, Maskarinec G, et al. Abstract P3-03-02: Can artificial intelligence derived ultrasound breast density provide comparable breast cancer risk estimates to density derived from mammograms. Cancer Research [Internet]. 2023 Mar 1 [cited 2023 Mar 15];83(5_Supplement):P3-03–02. Cite