Purchase this article with an account.
Jimmy Jiang Liu, Yanwu Xu, Jun Cheng, Zhang Zhuo, Damon W K Wong, Fengshou Yin, Tien Y Wong; Integration of multiple source data for glaucoma detection. Invest. Ophthalmol. Vis. Sci. 2014;55(13):4820.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To introduce an approach for the detection of glaucoma through the integration of data from multiple sources and to evaluate the performance of this approach.
We developed a multiple source fusion approach for the detection of glaucoma. This approach integrates information from demographic data, imaging information from retinal image features, and genetic information from genome single-nucleotide polymorphisms (SNPs) through the use of a multiple kernel learning framework. In multiple kernel learning, a combined kernel is formed from a weighted linear combination of base kernels instead of a single kernel. This framework is used to fuse the heterogenous data sources for the multi-source glaucoma detection approach. The framework is shown in Fig. 1. The area under the Receiver Operating Characteristic Curve (AUC) for glaucoma detection is used to evaluate our multi-source approach. As a comparison we have also included the detection performance from each of the individual data sources.
The approach was tested on a dataset from 2,258 individuals from the Singapore Malay Eye Study which includes patient demographic data, retinal images and SNP information. Glaucoma was diagnosed in 100 of the individuals from the dataset. The results show that the combination of all three sources achieves an AUC of 0.869, which is better than the result from any other individual data source or combination of any two sources.
An approach for the fusion of multiple data sources from demographic, imaging and genetic data for glaucoma detection is tested. The promising experimental results on a large dataset show the potential of multi-source integration for disease detection.
This PDF is available to Subscribers Only