{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,21]],"date-time":"2026-07-21T20:34:37Z","timestamp":1784666077207,"version":"3.55.0"},"reference-count":31,"publisher":"American Medical Association (AMA)","issue":"6","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["JAMA Netw Open"],"abstract":"Importance<\/jats:title>Abusive head trauma (AHT) in children is often missed in medical encounters, and retinal hemorrhage (RH) is considered strong evidence for AHT. Although head computed tomography (CT) is obtained routinely, all but exceptionally large RHs are undetectable on CT images in children.<\/jats:p><\/jats:sec>Objective<\/jats:title>To examine whether deep learning\u2013based image analysis can detect RH on pediatric head CT.<\/jats:p><\/jats:sec>Design, Setting, and Participants<\/jats:title>This diagnostic study included 301 patients diagnosed with AHT who underwent head CT and dilated fundoscopic examinations at a quaternary care children\u2019s hospital. The study assessed a deep learning model using axial slices from 218 segmented globes with RH and 384 globes without RH between May 1, 2007, and March 31, 2021. Two additional light gradient boosting machine (GBM) models were assessed: one that used demographic characteristics and common brain findings in AHT and another that combined the deep learning model\u2019s risk prediction plus the same demographic characteristics and brain findings.<\/jats:p><\/jats:sec>Main Outcomes and Measures<\/jats:title>Sensitivity (recall), specificity, precision, accuracy, F1 score, and area under the curve (AUC) for each model predicting the presence or absence of RH in globes were assessed. Globe regions that influenced the deep learning model predictions were visualized in saliency maps. The contributions of demographic and standard CT features were assessed by Shapley additive explanation.<\/jats:p><\/jats:sec>Results<\/jats:title>The final study population included 301 patients (187 [62.1%] male; median [range] age, 4.6 [0.1-35.8] months). A total of 120 patients (39.9%) had RH on fundoscopic examinations. The deep learning model performed as follows: sensitivity, 79.6%; specificity, 79.2%; positive predictive value (precision), 68.6%; negative predictive value, 87.1%; accuracy, 79.3%; F1 score, 73.7%; and AUC, 0.83 (95% CI, 0.75-0.91). The AUCs were 0.80 (95% CI, 0.69-0.91) for the general light GBM model and 0.86 (95% CI, 0.79-0.93) for the combined light GBM model. Sensitivities of all models were similar, whereas the specificities of the deep learning and combined light GBM models were higher than those of the light GBM model.<\/jats:p><\/jats:sec>Conclusions and Relevance<\/jats:title>The findings of this diagnostic study indicate that a deep learning\u2013based image analysis of globes on pediatric head CTs can predict the presence of RH. After prospective external validation, a deep learning model incorporated into CT image analysis software could calibrate clinical suspicion for AHT and provide decision support for which patients urgently need fundoscopic examinations.<\/jats:p><\/jats:sec>","DOI":"10.1001\/jamanetworkopen.2023.19420","type":"journal-article","created":{"date-parts":[[2023,6,23]],"date-time":"2023-06-23T23:38:53Z","timestamp":1687563533000},"page":"e2319420","source":"Crossref","is-referenced-by-count":10,"title":["Development of a Deep Learning Model for Retinal Hemorrhage Detection on Head Computed Tomography in Young Children"],"prefix":"10.1001","volume":"6","author":[{"given":"Fatma","family":"Gunturkun","sequence":"first","affiliation":[{"name":"Quantitative Sciences Unit, Department of Medicine, Stanford University, Palo Alto, California"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Berna","family":"Bakir-Batu","sequence":"additional","affiliation":[{"name":"Center for Biomedical Informatics, University of Tennessee Health Science Center, Memphis"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Adeel","family":"Siddiqui","sequence":"additional","affiliation":[{"name":"Department of Radiology, University of Tennessee Health Sciences Center, Memphis"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Karen","family":"Lakin","sequence":"additional","affiliation":[{"name":"Department of Pediatrics, Vanderbilt University Medical Center, Nashville, Tennessee"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mary E.","family":"Hoehn","sequence":"additional","affiliation":[{"name":"Department of Ophthalmology, University of Tennessee Health Sciences Center, Memphis"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Robert","family":"Vestal","sequence":"additional","affiliation":[{"name":"Department of Ophthalmology, University of Tennessee Health Sciences Center, Memphis"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Robert L.","family":"Davis","sequence":"additional","affiliation":[{"name":"Center for Biomedical Informatics, University of Tennessee Health Science Center, Memphis"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nadeem I.","family":"Shafi","sequence":"additional","affiliation":[{"name":"Department of Pediatrics, University of Tennessee Health Sciences Center, Memphis"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"10","published-online":{"date-parts":[[2023,6,22]]},"reference":[{"key":"zoi230590r1","doi-asserted-by":"publisher","first-page":"118","DOI":"10.3389\/fneur.2020.00118","article-title":"Advances and future directions of diagnosis and management of pediatric abusive head trauma: a review of the literature.","volume":"11","author":"Iqbal O\u2019Meara","year":"2020","journal-title":"Front Neurol"},{"key":"zoi230590r2","doi-asserted-by":"publisher","first-page":"36","DOI":"10.1016\/j.chiabu.2016.09.001","article-title":"Prior opportunities to identify abuse in children with abusive head trauma.","volume":"60","author":"Letson","year":"2016","journal-title":"Child Abuse Negl"},{"issue":"3","key":"zoi230590r3","doi-asserted-by":"publisher","first-page":"268","DOI":"10.1016\/j.jaapos.2009.03.005","article-title":"Odds of abuse associated with retinal hemorrhages in children suspected of child abuse.","volume":"13","author":"Binenbaum","year":"2009","journal-title":"J AAPOS"},{"issue":"4","key":"zoi230590r4","doi-asserted-by":"publisher","first-page":"226.e1","DOI":"10.1016\/j.jaapos.2020.03.012","article-title":"Nonaccidental trauma in pediatric patients: evidence-based screening criteria for ophthalmologic examination.","volume":"24","author":"Ip","year":"2020","journal-title":"J AAPOS"},{"issue":"5","key":"zoi230590r5","doi-asserted-by":"publisher","first-page":"1009","DOI":"10.1016\/j.jpedsurg.2020.07.031","article-title":"Catching the red eye: a retrospective review of factors associated with retinal hemorrhage in child physical abuse.","volume":"56","author":"Casar Berazaluce","year":"2021","journal-title":"J Pediatr Surg"},{"issue":"19","key":"zoi230590r6","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3390\/s20195611","article-title":"Accurate and efficient intracranial hemorrhage detection and subtype classification in 3D CT scans with convolutional and long short-term memory neural networks.","volume":"20","author":"Burduja","year":"2020","journal-title":"Sensors (Basel)"},{"issue":"11","key":"zoi230590r7","doi-asserted-by":"publisher","first-page":"6191","DOI":"10.1007\/s00330-019-06163-2","article-title":"Precise diagnosis of intracranial hemorrhage and subtypes using a three-dimensional joint convolutional and recurrent neural network.","volume":"29","author":"Ye","year":"2019","journal-title":"Eur Radiol"},{"issue":"5","key":"zoi230590r8","doi-asserted-by":"publisher","DOI":"10.3171\/2018.8.FOCUS17773","article-title":"Machine-learning analysis outperforms conventional statistical models and CT classification systems in predicting 6-month outcomes in pediatric patients sustaining traumatic brain injury.","volume":"45","author":"Hale","year":"2018","journal-title":"Neurosurg Focus"},{"issue":"6","key":"zoi230590r9","doi-asserted-by":"publisher","first-page":"1061","DOI":"10.1007\/s00247-021-05073-0","article-title":"Artificial intelligence in child abuse imaging.","volume":"51","author":"Sorensen","year":"2021","journal-title":"Pediatr Radiol"},{"issue":"6","key":"zoi230590r10","doi-asserted-by":"publisher","first-page":"1061","DOI":"10.3174\/ajnr.A6538","article-title":"Fully automated segmentation of globes for volume quantification in CT images of orbits using deep learning.","volume":"41","author":"Umapathy","year":"2020","journal-title":"AJNR Am J Neuroradiol"},{"issue":"1","key":"zoi230590r15","first-page":"6","article-title":"Comparison between XGBoost, LightGBM and CatBoost using a home credit dataset.","volume":"13","author":"Daoud","year":"2019","journal-title":"Int J Comput Info Engineer"},{"key":"zoi230590r16","doi-asserted-by":"publisher","article-title":"Explainable AI for trees: from local explanations to global understanding.","author":"Lundberg","DOI":"10.1038\/s42256-019-0138-9"},{"issue":"5","key":"zoi230590r17","doi-asserted-by":"publisher","first-page":"264","DOI":"10.2176\/nmc.ra.2015-0308","article-title":"Acute subdural hematoma in infants with abusive head trauma: a literature review.","volume":"56","author":"Karibe","year":"2016","journal-title":"Neurol Med Chir (Tokyo)"},{"issue":"5","key":"zoi230590r18","doi-asserted-by":"publisher","first-page":"983","DOI":"10.1016\/j.ophtha.2009.09.040","article-title":"A systematic review of the diagnostic accuracy of ocular signs in pediatric abusive head trauma.","volume":"117","author":"Bhardwaj","year":"2010","journal-title":"Ophthalmology"},{"issue":"2","key":"zoi230590r19","doi-asserted-by":"publisher","DOI":"10.1542\/peds.2018-1411","article-title":"The eye examination in the evaluation of child abuse.","volume":"142","author":"Christian","year":"2018","journal-title":"Pediatrics"},{"issue":"5","key":"zoi230590r20","doi-asserted-by":"publisher","first-page":"1172","DOI":"10.3171\/2014.2.JNS131248","article-title":"Comparison of CT and clinical findings of Terson\u2019s syndrome in 121 patients: a 1-year prospective study.","volume":"120","author":"Koskela","year":"2014","journal-title":"J Neurosurg"},{"issue":"6","key":"zoi230590r21","doi-asserted-by":"publisher","first-page":"1027","DOI":"10.1007\/s00701-016-2766-8","article-title":"Terson syndrome in aneurysmal subarachnoid hemorrhage-its relation to intracranial pressure, admission factors, and clinical outcome.","volume":"158","author":"Joswig","year":"2016","journal-title":"Acta Neurochir (Wien)"},{"issue":"2","key":"zoi230590r22","doi-asserted-by":"publisher","first-page":"247","DOI":"10.1111\/jon.12285","article-title":"Terson\u2019s syndrome: diagnostic comparison of ocular sonography and CT.","volume":"26","author":"B\u00e4uerle","year":"2016","journal-title":"J Neuroimaging"},{"issue":"9","key":"zoi230590r23","doi-asserted-by":"publisher","first-page":"1363","DOI":"10.1007\/s00247-015-3312-1","article-title":"MR detection of retinal hemorrhages: correlation with graded ophthalmologic exam.","volume":"45","author":"Beavers","year":"2015","journal-title":"Pediatr Radiol"},{"issue":"1","key":"zoi230590r24","doi-asserted-by":"publisher","first-page":"15","DOI":"10.1111\/jon.v30.1","article-title":"MRI findings in pediatric abusive head trauma: a review.","volume":"30","author":"Orman","year":"2020","journal-title":"J Neuroimaging"},{"issue":"2","key":"zoi230590r25","doi-asserted-by":"publisher","first-page":"527","DOI":"10.1016\/j.jpeds.2013.01.063","article-title":"Dedicated retinal examination in children evaluated for physical abuse without radiographically identified traumatic brain injury.","volume":"163","author":"Greiner","year":"2013","journal-title":"J Pediatr"},{"key":"zoi230590r26","doi-asserted-by":"publisher","first-page":"11","DOI":"10.1016\/j.pediatrneurol.2021.11.004","article-title":"Ocular injuries in pediatric patients admitted with abusive head trauma.","volume":"127","author":"Weiss","year":"2022","journal-title":"Pediatr Neurol"},{"issue":"4","key":"zoi230590r27","doi-asserted-by":"publisher","DOI":"10.1542\/peds.2020-0203","article-title":"Abusive head trauma in infants and children.","volume":"145","author":"Narang","year":"2020","journal-title":"Pediatrics"},{"issue":"1","key":"zoi230590r28","doi-asserted-by":"publisher","DOI":"10.1186\/s40537-019-0197-0","article-title":"A survey on image data augmentation for deep learning.","volume":"6","author":"Shorten","year":"2019","journal-title":"J Big Data"},{"issue":"9","key":"zoi230590r30","doi-asserted-by":"publisher","first-page":"665","DOI":"10.1097\/PEC.0000000000001605","article-title":"Retinal hemorrhages: abusive head trauma or not?","volume":"34","author":"Hansen","year":"2018","journal-title":"Pediatr Emerg Care"},{"issue":"2","key":"zoi230590r31","doi-asserted-by":"publisher","first-page":"131","DOI":"10.1016\/j.jaapos.2015.12.008","article-title":"The natural history of retinal hemorrhage in pediatric head trauma.","volume":"20","author":"Binenbaum","year":"2016","journal-title":"J AAPOS"},{"key":"zoi230590r11","unstructured":"Simonyan? K, Vedaldi? A, Zisserman? A. Deep inside convolutional networks: visualising image classification models and saliency maps.? arXiv. Preprint posted online December 20, 2013. doi:10.48550\/arXiv.1312.6034"},{"key":"zoi230590r12","doi-asserted-by":"crossref","unstructured":"Zeiler? MD, Fergus? R. Visualizing and understanding convolutional networks.? arXiv. Preprint posted online November 12, 2013. doi:10.1007\/978-3-319-10590-1_53?","DOI":"10.1007\/978-3-319-10590-1_53"},{"key":"zoi230590r13","unstructured":"Springenberg? JT, Dosovitskiy? A, Brox? T, Riedmiller? M. Striving for simplicity: the all convolutional net.? arXiv. Preprint posted online December 21, 2014. doi:10.48550\/arXiv.1412.6806"},{"key":"zoi230590r14","unstructured":"Ke? G, Meng? Q, Finley? T, . LightGBM: a highly efficient gradient boosting decision tree. Accessed May 15, 2023. https:\/\/github.com\/Microsoft\/LightGBM"},{"key":"zoi230590r29","unstructured":"Albus K. Technique parameters and anatomic coverage: CT\u2014pediatric chest module. Revised March 30, 2021. Accessed November 28, 2022. https:\/\/accreditationsupport.acr.org\/support\/solutions\/articles\/11000049641-technique-parameters-and-anatomic-coverage-ct-pediatric-chest-module-revised-3-30-2021-"}],"container-title":["JAMA Network Open"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/jamanetwork.com\/journals\/jamanetworkopen\/articlepdf\/2806474\/gunturkun_2023_oi_230590_1686842445.55938.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,6,23]],"date-time":"2023-06-23T23:39:01Z","timestamp":1687563541000},"score":1,"resource":{"primary":{"URL":"https:\/\/jamanetwork.com\/journals\/jamanetworkopen\/fullarticle\/2806474"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,6,22]]},"references-count":31,"journal-issue":{"issue":"6","published-print":{"date-parts":[[2023,6,1]]}},"URL":"https:\/\/doi.org\/10.1001\/jamanetworkopen.2023.19420","relation":{},"ISSN":["2574-3805"],"issn-type":[{"value":"2574-3805","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,6,22]]}}}