SHREVEPORT – The LSUS Artificial Intelligence and Machine Learning Lab captured a first-place award for its research poster on automatic spine segmentation at the national AIM-AHEAD annual meeting earlier this month.

The research, titled “SpineSegGAN: Automatic Segmentation of Lumbar Spine” detailed how their algorithms and model accurately segmented 93 percent of MRIs, which is an improvement compared to other models in existence (ranges from 75 to 80 percent).

Lab director Dr. Subhajit Chakrabarty guided a team of students from the Master of Science in Computer Systems Technology program in the research.

Student co-authors include Devesh Sarda, Udaysinh Rathod, and Mridula Mavuri.

“This award is significant because of the national scale of this conference,” Chakrabarty said of the event that draws more than 7,000 researchers and experts in artificial intelligence, machine learning and healthcare. “These are master’s students competing with and beating doctoral and postdoctoral researchers along with medical doctors and other faculty members who work in established medical research labs full-time.

“Our students take classes, so they are incredibly enthusiastic and highly motivated to perform this level of research in the time they do have.”

AIM-AHEAD is a grant program under the National Institutes of Health that empowers the development of AI/ML models to perform research in the medical field, starting with electronic health record data.

The LSUS team won in the category of hub-specific projects, one of 16 research categories that featured 220 total research posters at the Dallas-based conference.

“Everyone is coming up to our posters asking if we’re doing our Ph.D.’s on this – but we’re only master’s students,” Rathod said. “We really have to thank Dr. Chakrabarty because of his direction and guidance.

“He helps us format our papers and posters and puts the final touches that make it special.”

Spine segmentation is the process of precisely outlining anatomical structures, specifically vertebrae, intervertebral discs, and the spinal canal from medical images like MRI and CT scans.

This task, performed by radiologists, assists the patient’s medical team in diagnosing conditions, identifying abnormalities and planning surgical procedure paths.

“The model identifies and highlights spinal structures to where a doctor may be able to tell if the spine is constricted or can see discs moving or can see a hernia,” Sarda said. “We know radiologists are really swamped, so these types of AI tools can help doctors process more MRIs by validating rather than diagnosing.

“Doctors can validate the diagnosis, or if they notice something that is out of place or they don’t concur with the model’s diagnosis, they can take a deeper look at that particular case.”

The LSUS team trained its model on publicly available data and MRI images, allowing for the necessary volume and quality of data to produce accurate results.

Chakrabarty said the goal of AI/ML isn’t to replace doctors but to assist the health system in becoming more efficient and accurate.

“There’s a big value to real surgeons and radiologists when they can get feedback from an automated system in mere seconds,” Chakrabarty said. “The system can validate their thinking, and if there is disagreement, there can be more scrutiny applied.

“AI works well when the data challenges are large. Complex MRI imaging and databases with millions of data points, or something only experienced doctors would be qualified to do – that’s where AI can do extremely well.”

Not only did LSUS’s students need to be experts in AI/ML, they also needed to be well-versed in the lumbar spine region and the processes with which doctors use to make diagnoses and perform surgeries.

“That was a big piece of it, but the students took strongly to it and were enthused to learn about the medicine/biology part of it,” Chakrabarty said. “It’s significant additional work, but now when doctors are talking about what they’re doing and what they need, we know exactly what they are saying.”

The LSUS AI/ML lab presented a total of three research posters, but they walked away with more than just a first-place award.

Students were valued participants in group discussions stemming from speakers and other presentations.

“(Jiajie Zhang, dean of the UTHealth Houston School of Biomedical Informatics) is talking about how AI is geared more toward the younger generation, but that we can’t use AI as a substitution for our own knowledge,” Rathod said. “We should be validating what the AI is saying rather than just relying on it.”

Chakrabarty added that the speed of which LSUS students are implementing AI in their research is why they’re able to compete with more experienced researchers.

“Attending this conference is motivating for our students and allows our team to get a perspective about what direction this field is moving,” Chakrabarty said. “It’s also great name recognition for the university.

“We’ve received three offers to collaborate on grants from this conference, and others are asking a lot of questions about our students.”

The first-place poster is the fourth such honor for the lab, with the first two coming at the Louisiana Biomedical Research Network and the third at a previous AIM-AHEAD conference.

“We have momentum, and we want to keep that going,” Sarda said. “We’re being provided with a unique experience in that we’re able apply AI/ML and deep learning in real life.

“We’re able bring new perspectives as computer scientists and bring algorithms that can help real people in their lives.”

That type of recognition could help when the lab writes AIM-AHEAD grants, of which they’ve been awarded two in the recent past and are in contention for a third.

Grants fund student research positions to allow for further exploration in the field.