From pixels to predictions: Students can uncover AI uses in the lab

Mahsa Dabagh, professor of biomedical engineering, helps a male student as he sits in front of a computer looking at medical scans.
Mahsa Dabagh, associate professor, biomedical engineering, helps senior Miles Wehner decode slices of 3D tumor scans by labeling the visual features that indicate cancer. The two are now automating this time-consuming chore.

Cancer research often demands patience – hours of labeling 3D images, tracing tumor boundaries pixel by pixel before the real work can even begin. 

In Associate Professor Mahsa Dabagh’s lab, artificial intelligence is reshaping how researchers predict and treat disease. Dabagh, a biomedical engineering faculty member, studies how blood flows through and feeds tumors and how subtle biological and environmental cues can signal cancer’s next move. 

“AI helps us spend less time exploring the data and more time using it to design solutions,” Dabagh said. 

Making Waves of Impact
A newcomer to AI seizes the opportunity to fast-track the time-consuming task of segmentation.
a young man with blonde hair sitting behind a computer screen looks at the camera.
Biomedical engineering student Miles Smith started by manually labeling medical scans of tumors. Now he’s training AI to do the time-consuming job faster and smarter. ‘Learning these methods will be useful no matter where I end up,’ he said.

Her team is creating a deep-learning system that doesn’t just scan for tumors – it builds a virtual model of them. By analyzing MRI and biopsy images alongside detailed patient data such as protein activity, family medical history, and lifestyle factors, the tool can recognize not only what a tumor looks like, but how it might behave. 

“Research in the last decade shows that one data source is rarely enough for accurate prediction,” Dabagh said. “AI’s strength is in pulling together all those layers of information to create a more complete picture.” 

Undergraduate contributions 

Biomedical engineering senior Miles Wehner’s summer role in Dabagh’s lab was to work on the first step of the project, labeling tumors in MRI scans, a process that can take an hour per patient. Going slice by slice through hundreds of 3D images, Wehner had to carefully trace and label the tumor areas pixel by pixel.  

He soon saw the potential for AI to speed things up. Now he’s building a platform that uses both his own segmented scans and publicly available datasets to automate the entire process, bringing AI into the process earlier than the final predictions. 

“It’s still a lot of work to train our models,” Dabagh said, “but it’s worth it. It can improve clinicians’ predictions and give them personalized guidance for each patient’s treatment. That means saving more lives.” 

For Wehner, who only started coding through a summer program at UWM that introduces students to AI and programming, the research opened new doors. The program, Cyberinfrastructure Comprehensive, Applied and Tangible Summer School, or CIberCATSS, is funded by the National Science Foundation through 2026 and students with a faculty mentor can be admitted. 

“They say AI is the future. Learning these methods will be useful no matter where I end up,” he said. “Doing this research helped me see what’s possible.”