Varun Ranganathan, MC Optom

Clinical Optometrist

An OCULAR Interface Exclusive

Keywords: Recurrent Neural Networks, Amblyopia, Macular Degeneration, Glaucoma, Artificial Intelligence in Eyecare

Recurrent Neural Networks (RNNs) are a type of artificial neural network in which results of an output are fed back into the network, hence they are also feedback neural networks. By using the current output as an input for the next output, it can understand the context better¹. 

Both Convolutional Neural Network (CNNs) and RNNs are used broadly in eye-care and their usage depends on the type of data being analysed.  

RNNs are used in understanding time-sequential data and their usage in healthcare is especially important as it can help identify changes over time because of its ability in temporal dependencies².  

RNN in Glaucoma 

Glaucoma is characterised by progressive degeneration of optic nerve and loss of retinal ganglion cells which leads to impairment of visual field ³. Monitoring the condition’s progression is important for treatment options such as trabeculectomy, which significantly reduces the rate of progression ⁴. The combined model of CNN and RNN trained on fundus videos which contain temporal features has had better success in detecting glaucoma than the ones trained with only CNN on fundus images. Moreover, RNNs are more reliable in predicting future changes in visual field than other linear methods ⁵.  

RNN in Amblyopia 

Amblyopia is a visual disorder due to the failure of cortical visual development in one or both eyes due to ocular pathology early in life ⁶. The entire pathophysiology of amblyopia is still not fully understood, especially how higher visual areas are affected. Since RNNs can mimic the regions of higher visual cortex, they can be used as a viable option to understand amblyopia and thus aid in early detection.  

RNN in Macular Degeneration 

Age-related macular degeneration is a degenerative disease which can lead to progressive vision loss ⁷. Intermediate and late AMD can occur in both dry and wet subtypes and the rate of progression is at 28% in five years ⁸. The disease progression in both subtypes can be monitored using RNN and over a short timeframe this approach has been successful but over a long timeframe its efficacy has decreased ⁹. Though the prediction of progression of dry AMD is of little use as there is no treatment, it can help patients by equipping themselves better in the late stage. 

Prospects 

The use of RNN still needs more research to test its robustness in disease prediction and progression in other eye conditions. Its potential in empowering both clinicians and patients cannot be overstated.  

References 

1. N. Dey, S. Borra, A. S. Ashour, F. Shi. Machine Learning in Bio-Signal Analysis and Diagnostic Imaging, Academic Press, 2019, Page iv, ISBN 9780128160862, https://doi.org/10.1016/B978-0-12-816086-2.09993-8. 
2. S. K Zhou, D. Rueckert, G. Fichtinger. The Elsevier and MICCAI Society Book Series, Handbook of Medical Image Computing and Computer Assisted Intervention, Academic Press, 2020, Pages xvii-xxvi, ISBN 9780128161760, https://doi.org/10.1016/B978-0-12-816176-0.00005-3. 
3. Schuster AK, Erb C, Hoffmann EM, Dietlein T, Pfeiffer N. The Diagnosis and Treatment of Glaucoma. Dtsch Arztebl Int. 2020 Mar 27;117(13):225-234. doi: 10.3238/arztebl.2020.0225. PMID: 32343668; PMCID: PMC7196841. 
4. Koenig, S.F., Montesano, G., Fang, C.E.H. et al. Effect of trabeculectomy on the rate of progression of visual field damage. Eye 37, 2145–2150 (2023). https://doi.org/10.1038/s41433-022-02312-y. 
5. Keunheung Park; A Novel Visual Field Prediction Using Deep Learning: A Recurrent Neural Network Architecture. Invest. Ophthalmol. Vis. Sci. 2019;60(9):2438. 
6. Blair K, Cibis G, Gulani AC. Amblyopia. [Updated 2023 Aug 8]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2024 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK430890/. 
7. Heesterbeek TJ, Lorés-Motta L, Hoyng CB, Lechanteur YTE, den Hollander AI. Risk factors for progression of age-related macular degeneration. Ophthalmic Physiol Opt. 2020 Mar;40(2):140-170. doi: 10.1111/opo.12675. Epub 2020 Feb 25. PMID: 32100327; PMCID: PMC7155063. 
8. Stahl A. The Diagnosis and Treatment of Age-Related Macular Degeneration. Dtsch Arztebl Int. 2020 Jul 20;117(29-30):513-520. doi: 10.3238/arztebl.2020.0513. PMID: 33087239; PMCID: PMC7588619. 
9. Banerjee I, de Sisternes L, Hallak JA, Leng T, Osborne A, Rosenfeld PJ, Gregori G, Durbin M, Rubin D. Prediction of age-related macular degeneration disease using a sequential deep learning approach on longitudinal SD-OCT imaging biomarkers. Sci Rep. 2020 Sep 22;10(1):15434. doi: 10.1038/s41598-020-72359-y. PMID: 32963300; PMCID: PMC7508843.