Machine Learning Can Identify Eyes at Risk for Diabetic Retinopathy Progression
Automated machine learning models have shown promise in predicting the progression of diabetic retinopathy (DR) based on ultra-widefield retinal images, according to a recent study published in JAMA Ophthalmology. The findings suggest that this technology could help identify eyes that are at risk for DR progression, potentially leading to improved outcomes and reduced costs.
The study, led by Dr. Paolo S. Silva from Harvard University, involved the analysis of 1,179 deidentified ultra-widefield retinal images with mild or moderate nonproliferative DR (NPDR) over a three-year period. The researchers trained an automated machine learning model to predict the progression of DR using these images.
The results showed that the model had an area under the precision-recall curve of 0.717 for baseline mild NPDR and 0.863 for moderate NPDR. When validated, the model demonstrated a sensitivity of 0.72 and specificity of 0.63 for eyes with mild NPDR, with an overall accuracy of 64.3 percent. For eyes with moderate NPDR, the model achieved a sensitivity of 0.80, specificity of 0.72, and an accuracy of 73.8 percent.
In terms of identifying eyes that progressed two steps or more, the model performed well in the validation set. It successfully detected six out of nine eyes (75 percent) with mild NPDR and 35 out of 41 eyes (85 percent) with moderate NPDR that experienced progression within the follow-up period. Furthermore, the model identified all four eyes with mild NPDR that progressed within six months and one year, as well as eight out of nine (89 percent) with moderate NPDR that progressed within six months and 17 out of 20 (85 percent) that progressed within one year.
The authors of the study emphasize the potential of machine learning algorithms to refine the prediction of disease progression and identify individuals at the highest short-term risk. By doing so, the use of these algorithms could contribute to cost reduction and improve vision-related outcomes for patients with DR.
Nevertheless, it is worth noting that several authors involved in the study disclosed ties to Optos, a company specializing in retinal imaging. This disclosure ensures transparency and acknowledges any potential conflicts of interest.
In conclusion, the findings of this study highlight the potential of machine learning models that utilize ultra-widefield retinal images to predict the progression of diabetic retinopathy. Further research and validation are necessary to confirm the reliability and effectiveness of these models. Ultimately, if successfully implemented, this technology could play a crucial role in identifying patients at a higher risk of DR progression, allowing for early interventions and targeted treatments to preserve vision and prevent complications.
