A newly developed AI system from the University of Houston is aiming to replicate how expert radiologists examine chest X-rays, with potential implications for improving diagnostic accuracy and medical training.

The system, known as MedGaze, was developed by Hien Van Nguyen, associate professor of electrical and computer engineering, and detailed in a recently published Scientific Reports. MedGaze is trained to model where radiologists look, how long they fixate, and the sequence in which they examine chest X-rays, using data from thousands of recorded eye-tracking sessions.

“We're not just trying to guess what a radiologist will do next; we’re helping teach machines and future radiologists how to think more like experts by seeing the world as they do,” Nguyen said.

Described as a “digital gaze twin,” the software combines image data with radiology reports to predict where a radiologist is likely to look next. Its goal is twofold: support radiology education by modeling expert attention, and enhance AI diagnostic systems by guiding them to prioritize the same image regions as clinicians.

Nguyen emphasized that unlike conventional approaches in computer vision, which tend to focus on single-object detection, MedGaze is capable of modeling longer and more complex scan paths—necessary for identifying multiple abnormalities within a single image.

“Specifically, the key technical innovation of MedGaze is its capability to model fixation sequences that are an order of magnitude longer than those handled by the current state-of-the-art methods,” he said.

While the system is currently focused on chest X-rays, Nguyen’s team plans to extend the framework to other imaging modalities, including MR and CT, to support broader clinical applications.

MedGaze also offers workflow insights by identifying images that demand higher cognitive effort, potentially aiding resource planning in hospital radiology departments.