AI Tool Analyzes 30K Data Points Per Medical Imaging Pixel in Cancer Search

By LabMedica International staff writers
Posted on 16 Jan 2025

Image: The AI tool can search through data and histology images for much more precise information on cancer treatment effectiveness (Photo courtesy of Shutterstock)

A new artificial intelligence (AI)-powered tool can detect cell-level characteristics of cancer by analyzing data from very small tissue samples, some as tiny as 400 square micrometers, equivalent to the width of five human hairs. The tool, called MISO (Multi-modal Spatial Omics), processes vast amounts of data and applies insights to even the smallest regions on medical images. It has the potential to guide doctors toward the most effective therapies for various cancers, according to a recent paper about MISO published in Nature Methods.

MISO was developed by researchers at the Perelman School of Medicine at the University of Pennsylvania (Philadelphia, PA, USA) to work in the field of "spatial multi-omics." This area of research aims to gain insights into different conditions by considering the physical arrangement of tissue while examining various "-omics" modalities, such as transcriptomics (gene expression), proteomics (proteins), and metabolomics (metabolites and their processes), among others. In spatial transcriptomics, for example, a single pixel in an image contains 20,000 to 30,000 data points that need to be analyzed across multiple -omics layers, and this number can increase significantly if multiple omic layers are considered. By comparison, MRI and CT scans have only one data point (shades of gray) per pixel to interpret. Without AI tools like MISO, doctors and researchers would find it nearly impossible to uncover the valuable insights that the system can detect.

Using MISO, the researchers uncovered new information about several types of cancer, including bladder, gastric, and colorectal cancers, by analyzing data and images from donated patient tissue. In bladder cancer, MISO identified a specific group of cells responsible for forming tertiary lymphoid structures, which are associated with better responses to immunotherapy. In gastric cancer, MISO was able to differentiate cancer cells from the surrounding mucosa. In colorectal cancer, the system identified various sub-classes of cancer cells, shedding light on the distinct malignant cells within a single tumor. MISO was also used to analyze non-cancerous brain tissue structures.

These breakthroughs can guide more effective therapies, improve survival rates, and provide insights that are very challenging, if not impossible, to obtain without an advanced AI tool like MISO. Moving forward, the team aims to expand their knowledge of spatial -omics and pathology imaging to enhance MISO’s capabilities, including the ability to analyze multiple tissue samples simultaneously, which would greatly increase its output. While some data, such as epigenetic marks (chemicals that regulate DNA and are influenced by the environment), have not yet been widely measured, MISO’s AI system allows it to "learn" from the information it processes, enabling it to recognize new data as it becomes more available.

“As the field of spatial omics advances, it has become possible to measure multiple -omics modalities from the same tissue slice, providing complementary information and offering a more comprehensive, insightful view,” said Mingyao Li, PhD, the study’s senior author and a professor of Biostatistics and Digital Pathology. “MISO addresses a huge data challenge by enabling simultaneous analysis of all spatial -omics modalities, as well as microscopic anatomy images when available. It is the only method that is able to handle datasets like these with hundreds of thousands of cells per sample.”

Related Links:
Perelman School of Medicine