Artificial intelligence opens the door for greater precision and deeper understanding. That’s why Sadie Sims, a student at the University of Tennessee College of Veterinary Medicine, took a look at how artificial intelligence can serve as a tool in scoring lung lesions of pig lungs. Sims worked with Heather Kittrell, DVM, with Boehringer Ingelheim through the Swine Veterinary Internship Program at Iowa State University College of Veterinary Medicine.
“The current methodology to score porcine lungs is rather subjective as there are multiple methods available and different training levels/experience levels of practicing swine veterinarians,” Sims explained at the 2025 American Association of Swine Veterinarians (AASV) annual meeting. She was one of 15 students who competed in the Veterinary Student Poster Competition at the AASV annual meeting. Read more here.
Using images from previous clinical trials examining porcine reproductive and respiratory syndrome (PRRS) and Mycoplasma hyopneumoniae, Sims outlined the images using a program called LabelMe. This program helped differentiate the background the lungs were resting on at the time of imaging and to differentiate healthy lungs from the lung lesions.
“Dr. Shuo Wang took these images and performed random color, brightness, contrast and more to help conform all images to be as similar as possible as we were using two different sets of images from different years,” Sims says.
The dataset was split into a training dataset that contained 662 images and a testing dataset that contained 165 images. The training dataset was then run through a pyramid scene parsing network to segment the images to train the artificial intelligence model.
“An important thing to note is that the artificial intelligence model learns best with higher amounts of data to analyze,” she says. “The testing dataset was then used to evaluate the final model once it was completely trained.”
The testing dataset images will be segmented and analyzed via the artificial intelligence model for interpretation, Sims explains. The value used for analysis is called Intersection over Union (IOU). The goal is to be as close to 1 as possible but anything over >0.5 is considered satisfactory.
“Ultimately, the background and healthy lung were analyzed very successfully, with an IOU of 0.9806 and 0.9023, respectively,” she says. “However, the lesion only received an IOU of 0.2668, showing that the model was not as good at mapping the lesions on the lung as we had hoped. There was decreased detail in lesion recognition by the AI model compared to the outlined lesions analyzed by myself.”
The AI model did well at distinguishing between background and healthy lung but needs improvement when distinguishing between healthy and diseased lung.
Artificial Intelligence in the Pork Industry
Perhaps the most important discovery in Sims’ mind was the wide variety of uses for artificial intelligence that are possible in the pork industry.
“Artificial intelligence is constantly developing and improving,” she says. “I had no background in artificial intelligence prior to this project so I found it very interesting that it is being used to improve the veterinary medical profession. I also learned a lot about porcine lung pathology scoring and how important it would be to have a consistent, objective system.”
She believes the future uses of artificial intelligence in scoring lung lesions could have applications in slaughterhouses to pull carcasses off the line and help speed up the process in slaughterhouses. It could also be used in clinical research studies to provide consistency amongst research trials, she adds.
“I also think that (very far in the future), it may have applications to allow producers to perform the beginnings of the necropsy, take pictures of the lungs once removed, and have an application that could potentially tell the type of pathogen based off of the gross pathology,” Sims says. “This would allow the producer to take the information to their veterinarian and consult about future steps to protect the health of the barn. We all know swine veterinarians have to travel to farms far and wide to help provide medical care for producers and this could potentially help alleviate some of that stress.”
Your Next Read: Supplement Sows with Antioxidant Strategy to Move More Pigs Out the Door
