Vis enkel innførsel

dc.contributor.authorVoets, Mike
dc.contributor.authorMøllersen, Kajsa
dc.contributor.authorBongo, Lars Ailo
dc.date.accessioned2019-10-08T12:18:38Z
dc.date.available2019-10-08T12:18:38Z
dc.date.issued2019-06-06
dc.description.abstractWe have attempted to reproduce the results in <i>Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs</i>, published in JAMA 2016; 316(22), using publicly available data sets. We re-implemented the main method in the original study since the source code is not available. The original study used non-public fundus images from EyePACS and three hospitals in India for training. We used a different EyePACS data set from Kaggle. The original study used the benchmark data set Messidor-2 to evaluate the algorithm’s performance. We used another distribution of the Messidor-2 data set, since the original data set is no longer available. In the original study, ophthalmologists re-graded all images for diabetic retinopathy, macular edema, and image gradability. We have one diabetic retinopathy grade per image for our data sets, and we assessed image gradability ourselves. We were not able to reproduce the original study’s results with publicly available data. Our algorithm’s area under the receiver operating characteristic curve (AUC) of 0.951 (95% CI, 0.947-0.956) on the Kaggle EyePACS test set and 0.853 (95% CI, 0.835-0.871) on Messidor-2 did not come close to the reported AUC of 0.99 on both test sets in the original study. This may be caused by the use of a single grade per image, or different data. This study shows the challenges of reproducing deep learning method results, and the need for more replication and reproduction studies to validate deep learning methods, especially for medical image analysis.en_US
dc.description.sponsorshipPublication fund of UiT The Arctic University of Norwayen_US
dc.descriptionSource at <a href=https://doi.org/10.1371/journal.pone.0217541>https://doi.org/10.1371/journal.pone.0217541</a>.en_US
dc.identifier.citationVoets, M., Møllersen, K. & Bongo, L.A. (2019). Reproduction study using public data of: Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. <i>PLoS ONE, 14</i>(6), e0217541. https://doi.org/10.1371/journal.pone.0217541en_US
dc.identifier.cristinIDFRIDAID 1703436
dc.identifier.doi10.1371/journal.pone.0217541
dc.identifier.issn1932-6203
dc.identifier.urihttps://hdl.handle.net/10037/16359
dc.language.isoengen_US
dc.publisherPLOSen_US
dc.relation.isbasedon<p>Our code is open sourced under a MIT license: <a href=https://github.com/mikevoets/jama16-retina-replication>https://github.com/mikevoets/jama16-retina-replication</a>. Our EyePACS and Messidor-2 gradability grades are in the same repository. The data underlying the results presented in the study are available from: <p>1. Kaggle EyePACS data: <a href=https://www.kaggle.com/c/diabetic-retinopathy-detection>https://www.kaggle.com/c/diabetic-retinopathy-detection</a> (a free Kaggle account is required to access the data). <p>2. Messidor-2 data: <a href=https://medicine.uiowa.edu/eye/abramoff>https://medicine.uiowa.edu/eye/abramoff</a> (publicly available).en_US
dc.relation.journalPLoS ONE
dc.rights.accessRightsopenAccessen_US
dc.subjectVDP::Technology: 500::Medical technology: 620en_US
dc.subjectVDP::Teknologi: 500::Medisinsk teknologi: 620en_US
dc.subjectDiabetic retinopathyen_US
dc.subjectAlgorithmsen_US
dc.subjectDeep learningen_US
dc.subjectNeural networksen_US
dc.subjectOpen dataen_US
dc.subjectSource codeen_US
dc.subjectEdemaen_US
dc.subjectMachine learning algorithmsen_US
dc.titleReproduction study using public data of: Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographsen_US
dc.typeJournal articleen_US
dc.typeTidsskriftartikkelen_US
dc.typePeer revieweden_US


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel