The Speed of AI in Analyzing Heart Scans

According to a study conducted by researchers from UC San Francisco, advanced machine learning is faster, more accurate and more effective than board-certified echocardiographers. 

In their research, researchers used more than 180,000 echo images to train the computer to assess the most common echocardiogram views and they tested both the computer and the human technicians on new samples. After the test they found that computers were 91.7 to 97.8 percent accurate whereas humans were only accurate 70.2 to 83.5 percent. 

A senior Dr Rima Arnaout, a cardiologist at UCSF medical centre said, “This is providing a foundational step for analyzing echocardiograms in a comprehensive way”. Digital imaging is a vital part of medical diagnosis and interpreting medical images such as echocardiograms is complex and it takes too much time. Usually echo consists of various video clips, still images and heart recording from different angles or views. It makes it difficult for humans to offer accurate analyses. 

Researchers build a multilayer neural network and use supervised learning to classify 15 standard views. Randomly chose 80 percent of the images for training and reserved 20 percent for testing. 
The computer classified images from 12 video views with 97.8 percent accuracy whereas humans demonstrated accuracy is 70.2 to 83.5 percent. Dr Arnaout said, One of the drawbacks of multilayer neural networks is they need a lot of training data.