Title of Article: Machine Learning Approach for Intraocular Disease Prediction Based on Aqueous Humor Immune Mediator Profiles
What are the key takeaway points from this article?
Although the eye is immune-privileged, many ophthalmic diseases will result in the
production of inflammatory markers. Techniques such as flow cytometry can be used to identify inflammatory markers in various intraocular fluids, however, these findings can be difficult to interpret. A team from Tokyo attempted to use artificial intelligence to predict the diagnoses of 17 intraocular diseases and to identify the biomarkers that carried the most weight in predicting these diagnoses.
In this study, the authors compared various AI models to interpolate cytokine levels found in the aqueous humor of various intraocular diseases. They found the Random Forest machine learning algorithm to be most appropriate for this application, with highest accuracy for acute retinal necrosis, vitreoretinal lymphoma, and endophthalmitis. These diseases had an area under the receiver operating characteristic curve of 0.97, 0.97 and 0.98, respectively. Furthermore, they found the immune mediators MIG, IFN- γ, and IP-10 to be most important for predicting a diagnosis of acute retinal necrosis, along with IL-6, G-CSF, and IL-8 for endophthalmitis and IL-10, IP-10 and angiogenin for vitreoretinal lymphoma.
These findings are exciting for many reasons. For one, it demonstrates that AI using intraocular biomarkers levels has the potential to predict diseases in the field of ophthalmology. Further, we can use AI to see which immune mediators have the greatest predictive power and consequently, this may lead to the development of novel therapeutic targets. Lastly, this may lead to AI models being able to aid with disease prognostication, which could lead to a more personalized therapy for the patient. Ultimately, this has enormous potential to help with diagnostics, prognostication, and management, leading to improved patient-centered care.
Publication Date: August 2021
Reference:
Nezu, N., Usui, Y., Saito, A., Shimizu, H., Asakage, M., Yamakawa, N., Tsubota, K.,
Wakabayashi, Y., Narimatsu, A., Umazume, K., Maruyama, K., Sugimoto, M., Kuroda, M., &
Goto, H. (2021). Machine Learning Approach for Intraocular Disease Prediction Based on
Aqueous Humor Immune Mediator Profiles. Ophthalmology, 128(8), 1197–1208.
https://doi.org/10.1016/j.ophtha.2021.01.019
Summary By: Daniel Lamoureux