AI Achieves Near-Perfect Accuracy in Reading Medical Devices

Regular health monitoring for patients at home or in hospital can be a significant challenge, often requiring time-consuming manual checks. To reduce this labour, researchers have designed an automated system that uses artificial intelligence to read standard biomedical devices like blood pressure monitors and digital thermometers.

The innovative two-step process begins with an AI model called YOLOv8. It acts as the system's eyes, rapidly scanning an image to find the exact location of the digital display. This detection stage proved highly effective, showing a 99.5% accuracy rate in trials on custom datasets.

Once the screen is located, a second AI, a Convolutional Neural Network, takes over to perform the recognition. It accurately reads the specific numerical values displayed, achieving 99.7% accuracy across 1,000 test images. This precision could revolutionise remote patient care by automating data collection, with plans to include devices like heart rate monitors in the future.