Ensemble of deep convolutional neural networks using multi-spectral vision transformer for diabetic macular edema classification from multicolor image

Min Cui; Xinke Gao; Jianyi Song; Ping Zhang

doi:10.1117/12.3005817

10 October 2023 Ensemble of deep convolutional neural networks using multi-spectral vision transformer for diabetic macular edema classification from multicolor image

Min Cui, Xinke Gao, Jianyi Song, Ping Zhang

Author Affiliations +

Proceedings Volume 12799, Third International Conference on Advanced Algorithms and Signal Image Processing (AASIP 2023); 127995S (2023) https://doi.org/10.1117/12.3005817
Event: 3rd International Conference on Advanced Algorithms and Signal Image Processing (AASIP 2023), 2023, Kuala Lumpur, Malaysia

Abstract

Diabetic Macular Edema (DME) is a prevalent condition that threatens vision. Early and accurate detection of DME, coupled with appropriate treatment, can reduce the risk of blindness for patients. Multicolor image (MCI) facilitates the diagnosis of DME by providing multiple spectral images of retinal structures. Recent advances in deep learning methods for multimodal analysis have demonstrated superior performance over manual analysis. However, existing techniques for DME classification still suffer from low accuracy due to the insufficient utilization of MCI's properties. To address this issue, we propose a Multi-Spectral Vision Transformer (MSVT) model as a potential solution. Our model first resizes images from different spectra and then splits them into image patches. These patches are then projected into linear embeddings and added with positional embeddings. The Transformer encoder is then applied for feature extraction. The features of each spectrum are processed by a Multilayer Perceptron (MLP) head and then fused by a feature fusion module. Finally, the features are compressed and fed into a classifier for classification. We evaluate the empirical performance of the proposed algorithm on our in-house datasets. The classifier achieves a prediction accuracy of 0.935, sensitivity of 0.912, specificity of 0.924, and an AUC of 0.938 for predicting the DME status of MCIs.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Min Cui, Xinke Gao, Jianyi Song, and Ping Zhang "Ensemble of deep convolutional neural networks using multi-spectral vision transformer for diabetic macular edema classification from multicolor image", Proc. SPIE 12799, Third International Conference on Advanced Algorithms and Signal Image Processing (AASIP 2023), 127995S (10 October 2023); https://doi.org/10.1117/12.3005817

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available