AI framework enhances transparency in chest X-ray analysis

When interpreting medical images, trust in AI-assisted diagnoses is essential. “When people understand the reasoning process and limitations behind AI decisions, they are more likely to trust and embrace the technology,” said Ngan Le, a University of Arkansas assistant professor of computer science and computer engineering. 

ItpCtrl-AI, short for interpretable and controllable artificial intelligence, was developed to make AI-assisted diagnoses more understandable. The research findings were published in Artificial Intelligence in Medicine. 

The research team recorded how radiologists examine chest X-rays, tracking both their gaze direction and the duration they focused on specific areas. This data was then used to generate heat maps, which guide the AI in prioritizing key regions within an image. By mirroring human radiologists' focus areas, ItpCtrl-AI is designed to enhance diagnostic accuracy while ensuring that the reasoning behind its conclusions is transparent. 

This structured approach allows medical professionals to interpret AI-driven results more effectively and intervene when necessary. “If an AI medical assistant system diagnoses a condition, doctors need to understand why it made that decision to ensure it is reliable and aligns with medical expertise,” Le emphasized. 

A transparent AI system also addresses legal and ethical concerns by enabling medical professionals to take responsibility for diagnoses informed by AI assistance. “If we don’t know how a system is making decisions, it’s challenging to ensure it is fair, unbiased, or aligned with societal values,” said Le. Transparency in AI is particularly crucial in medical settings, as well as in other high-stakes areas such as self-driving technology and finance. 

Le and her research team, in collaboration with the MD Anderson Cancer Center in Houston, are currently working on refining ItpCtrl-AI to analyze more complex, three-dimensional CT scans. 

MEDICA-tradefair.com; Source: University of Arkansas