Deep learning to diagnose Peripapillary Atrophy in retinal images along with statistical features. (February 2021)

• Record Type: Journal Article
• Authors: Sharma, Ambika; Agrawal, Monika; Dutta Roy, Sumantra; Gupta, Vivek; Vashisht, Praveen; Sidhu, Talvir
• Is Part Of: Biomedical signal processing and control. Volume 64(2021)
• Publication Date: 2021-02
• Subjects: Retinal images -- Peripapillary Atrophy -- Glaucoma diagnosis -- Deep learning CNN models -- Image processing; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.102254
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract: Peripapillary Atrophy (PPA, hereafter) is one of the major indicators of an irreversible eye disease named Glaucoma. An early detection of PPA is vital to avoid vision reduction caused by pathological myopia, or a permanent loss caused by Glaucoma. PPA is a pigmented crescent-shaped abnormality around the optic disc region. In this paper, we propose a fusion method to detect the atrophy by combining ResNet50-based deep features along with clinically significant statistical features of the region of interest (containing PPA). We show results of extensive experimentation with six publicly available databases, on which the system is also trained. The testing is on a rather difficult dataset of community camp-based images captured under poor lighting conditions with hand-held low-resolution ophthalmoscopes. We show encouraging experimental results of the combination of the generalization power of deep features and the medical science behind clinical hand-crafted features. Such a feature combination out-performs any one of the modalities in the difficult experimental set. We compare our results with the state-of-the-art in the area. The proposed method outperforms existing methods with average sensitivity, specificity and accuracy values of 95.83% each. To the best of our knowledge, this is the best accuracy reported in the literature, on large and varied datasets. Highlights: The methodology proposes a fusion-based method that uses a deep CNN as a feature extractor in combination with clinically significant hand-crafted statistical features. The work has been validated on community camp-based images captured under poor lightning conditions, poor sitting setups and low resolution hand-held ophthalmoscopes. Inter-dataset PPA variability is a key contribution of the work by enhancing the data variance and generalization of the proposed network. The proposed method is not particularly sensitive to the size, orientation and colour of PPA i.e., it works with various sizes and levels of PPA from small to large and low to severe respectively. We have evaluated the effectiveness and generalization capability of the proposed network on seven different datasets and achieves state-of-the-art detection performance, with the average detection accuracy as 95.83%.