AI Model Identifies Signs of Disease Faster and More Accurately Than Humans

Traditionally, researchers and medical professionals identify pathology, or signs of disease, by meticulously examining and annotating tissue samples under a microscope, a process that can take hours for each slide or image. Now, a "deep learning" artificial intelligence (AI) model can identify pathology in animal and human tissue images much more rapidly, and in many cases, with greater accuracy than humans. This development, outlined in Scientific Reports, could significantly accelerate disease-related research and holds promise for enhancing medical diagnoses, such as identifying cancer from biopsy images in just minutes.

To create the AI model, computer scientists at Washington State University (Pullman, WA, USA) trained it using images from prior epigenetic studies conducted by their team. These studies focused on molecular-level disease markers in tissues from rats and mice, including kidney, testes, ovarian, and prostate tissues. The researchers then tested the AI on images from additional studies, including those identifying breast cancer and lymph node metastasis. They discovered that the new AI model not only identified pathologies quickly but also did so faster than previous models, and in some cases, it detected instances that a trained human team had missed.

In epigenetic research, which examines changes to molecular processes influencing gene activity without altering the DNA itself, analysis can take years for large-scale studies. However, with the new AI model, the same data can be processed in just a few weeks. Deep learning is an advanced AI approach designed to replicate the human brain, surpassing traditional machine learning methods. This model is structured with a network of neurons and synapses. When the model makes an error, it "learns" from it using backpropagation, a technique that adjusts the network to correct the mistake, preventing it from happening again.

The research team developed the WSU deep learning model to process high-resolution, gigapixel images containing billions of pixels. To handle the large file sizes, which can slow down even powerful computers, the model analyzes smaller tiles of the image while maintaining their context within larger sections at lower resolution, similar to zooming in and out with a microscope. This model has already caught the attention of other researchers, with its potential to advance both research and diagnosis, especially in areas such as cancer and gene-related diseases. By using annotated images, such as those identifying cancer in tissue samples, researchers could train the AI model to perform similar tasks in medical settings.

“This AI-based deep learning program was very, very accurate at looking at these tissues,” said Michael Skinner, a WSU biologist and co-corresponding author on the paper. “It could revolutionize this type of medicine for both animals and humans, essentially better facilitating these kinds of analysis.”

Visit expo >

Gold Member

Flocked Fiber Swabs

Puritan® Patented HydraFlock®

Verification Panels for Assay Development & QC
Seroconversion Panels

New

TORCH Infections Test

TORCH Panel

New

Centromere B Assay

Centromere B Test