Blended Multi-Modal Deep ConvNet Features for Diabetic Retinopathy Severity Prediction

TL;DR

This paper proposes a multi-modal feature fusion approach using pre-trained ConvNet models to improve diabetic retinopathy severity prediction, achieving high accuracy on a benchmark dataset.

Contribution

It introduces a novel multi-modal fusion module combining features from multiple ConvNets, enhancing retinal image representation for better DR classification.

Findings

Fusion of features from Xception and VGG16 yields best results.

The proposed model achieves 97.41% accuracy in DR identification.

Blended features enable faster convergence of DNN with dropout.

Abstract

Diabetic Retinopathy (DR) is one of the major causes of visual impairment and blindness across the world. It is usually found in patients who suffer from diabetes for a long period. The major focus of this work is to derive optimal representation of retinal images that further helps to improve the performance of DR recognition models. To extract optimal representation, features extracted from multiple pre-trained ConvNet models are blended using proposed multi-modal fusion module. These final representations are used to train a Deep Neural Network (DNN) used for DR identification and severity level prediction. As each ConvNet extracts different features, fusing them using 1D pooling and cross pooling leads to better representation than using features extracted from a single ConvNet. Experimental studies on benchmark Kaggle APTOS 2019 contest dataset reveals that the model trained on proposed blended feature representations is superior to the existing methods. In addition, we notice that cross average pooling based fusion of features from Xception and VGG16 is the most appropriate for DR recognition. With the proposed model, we achieve an accuracy of 97.41%, and a kappa statistic of 94.82 for DR identification and an accuracy of 81.7% and a kappa statistic of 71.1% for severity level prediction. Another interesting observation is that DNN with dropout at input layer converges more quickly when trained using blended features, compared to the same model trained using uni-modal deep features.