2018
DOI: 10.1016/j.cell.2018.02.013
|View full text |Cite
|
Sign up to set email alerts
|

Learning from Everyday Images Enables Expert-like Diagnosis of Retinal Diseases

Abstract: Kermany et al. report an application of a neural network trained on millions of everyday images to a database of thousands of retinal tomography images that they gathered and expert labeled, resulting in a rapid and accurate diagnosis of retinal diseases.

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

0
16
0

Year Published

2018
2018
2023
2023

Publication Types

Select...
6
2

Relationship

0
8

Authors

Journals

citations
Cited by 24 publications
(16 citation statements)
references
References 7 publications
(13 reference statements)
0
16
0
Order By: Relevance
“…Reasons to learn features from raw data include that doing so often substantially improves performance ( 13 , 25 , 31 ); because such features can be transferred to other domains with small datasets ( 32 , 33 ); because it is time-consuming to manually design features; and because a general algorithm that learns features automatically can improve performance on very different types of data [e.g., sound ( 20 , 34 ) and text ( 23 , 35 )], increasing the impact of the approach. However, an additional benefit to deep learning is that if hand-designed features are thought to be useful, they can be included as well in case they improve performance ( 36 40 ).…”
Section: Background and Related Workmentioning
confidence: 99%
“…Reasons to learn features from raw data include that doing so often substantially improves performance ( 13 , 25 , 31 ); because such features can be transferred to other domains with small datasets ( 32 , 33 ); because it is time-consuming to manually design features; and because a general algorithm that learns features automatically can improve performance on very different types of data [e.g., sound ( 20 , 34 ) and text ( 23 , 35 )], increasing the impact of the approach. However, an additional benefit to deep learning is that if hand-designed features are thought to be useful, they can be included as well in case they improve performance ( 36 40 ).…”
Section: Background and Related Workmentioning
confidence: 99%
“…Advances in imaging technology, together with artificial intelligence (Bi et al 2019), have allowed researchers to create various diagnostic and treatment models and improved the diagnostic efficacy in liver cancer, dermatology, ophthalmology, lung and breast cancers, neurology, cardiovascular diseases, gastrointestinal endoscopy, and genetic diseases, etc. (Attia et al 2019;Chilamkurthy et al 2018;Coudray et al 2018;Esteva et al 2017;Gurovich et al 2019;Kermany et al 2018;Mori et al 2018;Rampasek and Goldenberg 2018;Yasaka et al 2018;Zou et al 2019). The purpose of the current study is to develop models using eXtreme Gradient Boosting (XGBoost) and deep learning to provide a preoperative non-invasive assessment method for MVI in HCC patients.…”
Section: Introductionmentioning
confidence: 99%
“…A type of deep learning model called the convolutional neural network (CNN) can be applied to identify objects and settings present in the image or make other image-related predictions. Typically, CNNs are initially trained with everyday images, 17,18 but they have been successfully repurposed for clinical applications including identifying diabetic retinopathy 19,20 and skin cancer. 21 Moreover, computationally efficient CNNs (ie, with fewer parameters) have now been developed for mobile devices, allowing images to be rapidly analyzed with a smartphone or other device without substantially compromising performance.…”
Section: Introductionmentioning
confidence: 99%