Researchers wanted to see if artificial intelligence (AI) software could help radiologists analyze breast ultrasound images. They tested a commercial AI program called Vis-BUS. Six radiologists examined 258 breast ultrasound images, half showing cancerous lesions and half showing benign ones. They read the images once without AI help and once with it, with a break in between to avoid memory bias.
When radiologists used the AI software, one technical measure of performance improved slightly. Their reading time also dropped from about 6 seconds to 3 seconds per case. However, key measures of diagnostic accuracy—like how often they correctly identified cancer or ruled it out—did not change significantly with the AI.
This was a small, early study that looked back at existing images. It did not test the AI in a real clinical setting where doctors make immediate decisions. The study also did not report any safety issues or conflicts of interest. The results suggest AI might help speed up analysis, but it did not clearly make radiologists more accurate in this specific test. More research in real-world settings is needed to see if this tool is truly helpful for patients.