2023
DOI: 10.1038/s41746-023-00748-4
|View full text |Cite
|
Sign up to set email alerts
|

Multimodal deep learning of fundus abnormalities and traditional risk factors for cardiovascular risk prediction

Abstract: Cardiovascular disease (CVD), the leading cause of death globally, is associated with complicated underlying risk factors. We develop an artificial intelligence model to identify CVD using multimodal data, including clinical risk factors and fundus photographs from the Samsung Medical Center (SMC) for development and internal validation and from the UK Biobank for external validation. The multimodal model achieves an area under the receiver operating characteristic curve (AUROC) of 0.781 (95% confidence interv… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
5
0

Year Published

2023
2023
2024
2024

Publication Types

Select...
6
1
1

Relationship

0
8

Authors

Journals

citations
Cited by 28 publications
(6 citation statements)
references
References 40 publications
0
5
0
Order By: Relevance
“…In the hybrid system proposed by Diaz-Pinto et al [ 24 ], retinal images and relevant clinical metadata were analyzed together to estimate cardiac indices—including left ventricular mass and left ventricular end-diastolic volume—and to predict incident myocardial infarction; this model showed an AUC of 0.80. In another study, a multimodal deep learning system was proposed by Lee et al [ 25 ] to predict current CVD and assess its performance against a clinical risk factors data-input model simultaneously. The authors found that the integrated network, which combined two networks using two different modalities and made full use of clinical metadata and fundus images, achieved the best performance, with an AUC of 0.872 in the external validation set.…”
Section: Resultsmentioning
confidence: 99%
See 1 more Smart Citation
“…In the hybrid system proposed by Diaz-Pinto et al [ 24 ], retinal images and relevant clinical metadata were analyzed together to estimate cardiac indices—including left ventricular mass and left ventricular end-diastolic volume—and to predict incident myocardial infarction; this model showed an AUC of 0.80. In another study, a multimodal deep learning system was proposed by Lee et al [ 25 ] to predict current CVD and assess its performance against a clinical risk factors data-input model simultaneously. The authors found that the integrated network, which combined two networks using two different modalities and made full use of clinical metadata and fundus images, achieved the best performance, with an AUC of 0.872 in the external validation set.…”
Section: Resultsmentioning
confidence: 99%
“…Therefore, multimodal and multitask learning architecture should be encouraged to isolate the alternations caused by confounding factors. Furthermore, the integrated input of clinical metadata and ocular images, in most cases, may improve predictive accuracy [ 25 ].…”
Section: Discussionmentioning
confidence: 99%
“…33 In a case-control study, Lee et al predicted the prevalence of CVD using a multimodal deep learning model integrating retinal images and traditional risk factors and compared the results to the predictive ability of PCE-ASCVD. 10 They found major performance improvement by using the deep learning model.…”
Section: Discussionmentioning
confidence: 99%
“…Since the review was published, we have identified several relevant articles on the topic. 10,11 Finally, we aim to evaluate the studies from a more clinical perspective (e.g. performance comparison with or added value to existing risk scores).…”
Section: Introductionmentioning
confidence: 99%
“…Another application of AI analysis of fundus images is for the prediction of systemic disorders, specifically cardiovascular disease risk. [ 2 ] Future AI models that collectively analyze fundus imaging, basic cardiovascular examination data, and simple investigations such as ECG may turn out to be good predictors of cardiovascular disease risk.…”
mentioning
confidence: 99%