Scientist develops fast and accurate AI cardiology tool

But it's not ready to replace doctors just yet, admits researcher.

A scientist in the US has developed a "fast" and "accurate" artificial intelligence system that can classify echocardiogram results using deep-learning algorithms.

Rima Arnaout, from the University of California San Francisco, said the AI system is capable of analysing heart scans faster and more accurately than human cardiologists.

However, speaking to IEEE Spectrum, she warned that the technology won't replace doctors, but could help double-check scans, speed-up evaluation and, perhaps, highlight factors the doctors may have missed. Currently, the tool is limited because it can only evaluate echocardiograms - ultrasounds of the heart.

"The best technique is still inside the head of the trained echocardiographer," she said.

As part of a recent study, Arnout got the AI system and professional cardiologists to sort through a plethora of heart images. The computer program was 92 per cent correct, while the humans had an accuracy rate of 79 per cent.

The researcher has published her findings in the academic journal Digital Medicine. Arnaout explains that she used deep learning algorithms and a so-called convolutional neural network to get an AI system to evaluate echocardiograms.

Analysing these images is not an easy task. Because the heart is complex in structure, echocardiographers must record the organ from several angles. In addition, as echocardiograms are based on ultrasound the images are far from clear and require a trained eye to read them.

The more complex aspect is working out which part of the heart these videos show, and that is where deep learning and artificial intelligence can be helpful.

"Imaging is a critical part of medical diagnosis. Interpreting medical images typically requires extensive training and practice and is a complex and time-intensive process," wrote Arnaout in the study.

"Deep learning, specifically using convolutional neural networks (CNNs), is a cutting-edge machine learning technique that has proven "unreasonably" successful at learning patterns in images and has shown great promise helping experts with image-based diagnosis in radiology, pathology and dermatology."

When it comes to analysing heart scans, cardiologists will often have access to high-resolution videos. However, Arnaout's AI tool was only able to process still images which were 60x80 pixels in size.

Despite the fact that the AI tool had the harder task, it still trumped the human experts. In an interview with IEEE Spectrum, Arnaout said: "These were excellent echocardiographers. But it's a hard task. We're not used to seeing the images shrunken down and out of context."