After self-driving cars, voice assistance and visual recognition, artificial intelligence could revolutionise another aspect of our lives: medicine. Thanks to its uncommon faculties, particularly machine learning, AI could one day have the power to predict illness well before any symptoms show. Some developers are already announcing their foray into what promises to be a colossal market.
Look me in the eyes, and I will diagnose
Google’s beacon AI, DeepMind, is certainly one of the most fascinating AI created to date. Proof of this statement lies in its latest act of prowess: the detection of a potential heart problem via a retina scan which was reported in Nature’s end of February review. By reading in between the lines of ocular scintigraphy, the Verily algorithm used by the Alphabet medical associate can predict the age, sex and blood pressure of the patient. What is more, the procedure is not invasive.
Analysing neurons to predict Alzheimer’s
In September of 2017, researchers from the Bari university in Italy, published a promising article on arXiv.org. The article mentions the development of an artificial intelligence able to detect cerebral changes caused by Alzheimer’s 10 years before they happen! To do this, the AI excels in the analysis of MRI scans to focalise on neuronal connectivity. Out of 148 scanners, the algorithm was able to successfully diagnose Alzheimer’s 86% of the time. What is more, it was revealed that this technology was able to detect the deficiency even in its early stages, with a success rate of 84%.
As soon as birth
Barbara Engelhardt is a computer science lecturer at Princeton University. She believes that finely combing everyone’s genetic code (thanks to AI) using machine learning could allow us to find hidden information about each individual. On October 2017, she explained in Nature being able to determine how genetic mutations are linked to the regulations of genes on different chromosomes in 44 human tissues. Thanks to machine learning and data analysis, she was also able to build a model capable of telling doctors and nurses if they should remove or keep a patient’s respiratory assistance.
Using deep learning to prevent blindness
The DreamUp Vision start-up developed an algorithm able to prematurely diagnose diabetic retinopathy, blindness caused by the degeneration of diabetics. This is especially important as this consequence of diabetes is not inevitable: if the patient is treated during the early stages, it is possible to avoid blindness completely. To do this, the algorithm uses retinography images of the back of the eye to detect stains or any signs that show arterial deterioration. If the AI is able to do this, it is because it has trained. In total, the DreamUp Vision algorithm has had to go through 90,000 retinal scans in order to reach optimal results. We can imagine this type of prowess being extended to all other medical fields.