Unleashing the power of retinal images to improve patients’ lives
By Jesica Levingston Mac Leod
“The eyes are the window to the soul” is turning from a metaphor to a reality, thanks to the recent developments in image analysis. This old fashion expression implies that the eyes of a person are supposed to give you insight into their inner world, like their thoughts and their health state. Due to the power of innovative image analysis programs the latest is turning into a fact. Indeed, every day more companies add their names to the long list of enterprises on the pursuit of the program that can predict different health features. Google has been working on pattern recognition in ocular images for a series of pathologies, like eye related diseases and smoking status. Recently, Google’ researchers, leaded by Dr. Lily Peng, predicted with 70% accuracy the likelihood for a person to suffer a heart attack within 5 years, only by studying images of the patients’ retina. The Holy Grail of Artificial intelligence (AI) was used for the program that gave Google already amazing results in the “heart status” prediction. They have published the findings in the Journal “Nature Biomedical Engineering”, demonstrating that their AI based method is as accurate on predicting heart disease than more invasive procedures, like blood test to measure cholesterol levels. Although, the deep learning strategy didn’t outperform the existing methods, this technology might help the doctors on the decision-making.
The prediction program was trained with data from 284.335 patients and validated on two independent data sets of 12.026 and 999 patients. The researchers are aware that a bigger data set is needed for training and validation. The readout or outcome from the program is a heatmap showing which pixels in an image are most important to predict the patient’s risks, for example based on the state of the blood vessels in the retina.
Braviithi shares the same aim of improving the diagnostic process and speeding treatment pathways, using the retinal fundus images for disease prediction. Braviithi’s technologies have shown around 88% accuracy in predicting diabetic retinopathy. Diabetic retinopathy is a diabetes complication, caused by damage to the blood vessels of the retina, which can produce blindness. However, the selected approach is not based on AI, but in a mathematical model that allows a broader option of image types than the AI methods, plus other advantages that mark this cutting-edge technology as a preferable prediction tool in areas where different machines and tools are used for generating the retinal images. Making use of innate characteristics of different features in the retina pre and post disease onset, Braviithi tool modeled them accordingly using shape, size, color, intensity, density, orientation and compactness attributes.
The quality control of the images that are going to be analyzed is another factor that shall be consider, and on this topic again Braviithi tool can be handier for the medical professionals to support the decision making.
Let’s keep our eyes open for the encouraging advances in image analysis that can help us detect health issues before it is too late.
Link to the article: https://www.nature.com/articles/s41551-018-0195-0