Transparent AI platform shows radiologists its decision-making blueprint for diagnosing breast cancer

A new artificial intelligence platform that detects cancerous lesions on mammograms has been developed to mimic the interpretation processes of radiologists. 

What's more, the algorithm, developed by radiologists and computer engineers at Duke University, lays out a blueprint that explains what image factors led to its conclusion.  

“We need algorithms that not only work but explain themselves and show examples of what they’re basing their conclusions on. That way, whether a physician agrees with the outcome or not, the AI is helping to make better decisions,” Joseph Lo, PhD, professor of radiology at Duke, said in a statement.

In order for the AI to read like a radiologist, it had to first be trained like a radiologist. It was first taught to identify suspicious lesions and to then disregard all surrounding tissue. Next, radiologists carefully labeled the mammographic lesions in a way that would allow the AI to focus on the lesion’s edges for abnormalities, which is what radiologists look for once a mass is identified.  

The AI was presented with 1,136 images of 484 patients with known cancerous and benign lesions. This helped the algorithm to recognize what margin abnormalities on mammograms look like and flag images that might merit additional investigation. 

“Other AIs are not trying to imitate radiologists; they’re coming up with their own methods for answering the question that are often not helpful or, in some cases, depend on flawed reasoning processes,” Alina Barnett, computer science PhD candidate at Duke and first author of the study, explained. 

When the AI’s performance was compared to radiologist interpretations, it did not outperform the human readers. However, its performance did as well as other algorithms and also offered the ability for radiologists to map out its decision-making process, which would allow them to pinpoint exactly where the interpretation went wrong. 

“This is what transparency in medical imaging AI could look like and what those in the medical field should be demanding for any radiology challenge,” insisted Cynthia Rudin, professor of electrical and computer engineering and computer science at Duke. 

The Duke MEDx High-Risk High-Impact Award was recently given to the experts to assist them in their continued research. The team hopes to forge forward with their efforts by next introducing additional physical characteristics, like lesion shape, to the algorithm. 

You can view the detailed research in Nature Machine Intelligence

Hannah murhphy headshot

In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.

Trimed Popup
Trimed Popup