DeepMind, a British Artificial Intelligence (AI) company owned by Google’s parent company Alphabet Inc., has successfully developed a system that can analyse retina scans and can spot symptoms of more than 50 sight-threatening eye diseases. Researchers from DeepMind, Moorfields Eye Hospital and University College London have used deep learning methods to create this intelligent software. The research published in the journal Nature Medicine, details how deep learning architecture is utilized to understand a set of three-dimensional optical coherence tomography (OCT) scans from patients.
The new AI system identifies eye diseases from 3D scans and then recommends the patient for treatment. At present, it can detect around 50 eye diseases, and researchers claim that the accuracy rate for detection of diseases is 94%, which they say is at par with or even better than world-leading eye specialists. The system has been trained on 14,884 scans from 7,621 patients, thus far.
The AI system adopts a human-like logical approach for analysing the highly complex OCT scans of patient retinas. Subsequently, the system assesses the 3D tissue map and makes decisions about what the diseases might be and how urgent they are for referral and treatment. Patients with sight problems are largely categorized into urgent, semi-urgent, routine and observation-only clinical categories. The tissue scanning stage is what is unique to this AI system—the other AI systems deployed thus far would read the OCT scans and pronounce a diagnosis, but the DeepMind AI systems allows clinicians to look at the 3D tissue map as well for detailed diagnosis.
The OCT scan for each eye typically takes around 10 minutes, with the use of infrared light bouncing off the surfaces of the eye—this is to create a 3D image of the tissues in the eye. The OCT scans are crucial when it comes to detecting, identifying and treating eye diseases—some of which could have had grave consequences later if not treated in time.
At present, this DeepMind AI system will undergo further testing and subsequently pass clinical trials and regulatory approvals, before it can be deployed in clinics and hospitals. Nevertheless, it is still quite unlikely that a patient would purely trust a machine, no matter how intelligent, with their treatment. As things are, AI systems such as the one DeepMind is developing here, will further assist the human doctors in diagnosis, deeper analysis and hopefully more accurate treatments.
Comments
0 comment