A team of scientists led by Sebastian Massaro from the University of Surrey (UK) taught a neural network to recognize signs of heart failure in patients' ECGs. The accuracy and speed of the diagnostic system is striking - it makes a 100% accurate diagnosis in just one session. Moreover, AI can also determine morphological features that indicate the severity of the disease.
It should be noted that statements about 100% accuracy should be taken critically so far. At the current stage, the AI has been trained on thousands of ECGs of healthy people and patients with severe deficiency. And therefore, the neural network will absolutely definitely distinguish a healthy person from a sick person, but what about those whose disease has just begun to develop, and the symptoms are not expressed? The problem is that it is much more difficult to find such patients and get their ECG, and therefore there is simply nothing to train the neural network on.
Massaro himself believes that hundreds and thousands of tests are needed before the AI diagnostician can be involved in real work. But, strictly speaking, it is not intended for this. Neurity is optimized specifically for high-speed analysis using small amounts of data that even a small wearable gadget can collect. This is to save people from having to undergo a long and tedious examination in a clinical setting. Instead, compact electronics will constantly monitor their health.
The devices for collecting bio-data already exist and are available to a wide range of users. But AI for analyzing the collected information is still in short supply, which is a bottleneck in digital diagnostics. The development of Massaro, like the projects of his colleagues from other countries, will help solve this problem.