Cross-validated radiomic SuStaIn identifies prognostically and transcriptomically distinct event-ordering phenotypes in diffuse glioma

Brain tumors are among the most complex and dangerous diseases a person can face. Among these, diffuse gliomas are a group of cancers that grow within the brain's soft tissue, spreading out like roots through a garden rather than forming a single, contained lump. For decades, doctors have tried to sort these tumors into categories to predict how they will behave and how a patient will respond to treatment. The most recent system relies on looking at the tumor's cells under a microscope and analyzing its genetic code, specifically checking for certain mutations in the DNA. While this molecular approach has improved diagnosis, it does not tell the whole story. Two patients with tumors that look identical under a microscope and share the same genetic markers can still have very different outcomes; one might survive for many years while the other declines rapidly. This gap suggests that there are hidden differences within the tumors that current tests cannot see.

To find these hidden differences, researchers are turning to magnetic resonance imaging, or MRI, which creates detailed pictures of the brain without needing surgery. Instead of just looking at these images with their eyes, scientists use a method called radiomics. This process turns the visual information in an MRI scan into thousands of tiny, precise numbers that describe the shape, texture, and brightness of the tumor. These numbers can reveal patterns of organization inside the tumor that are too subtle for the human eye to detect. The challenge has been figuring out how to use these numbers to understand the disease better. Most previous studies have tried to use these numbers to simply predict a known label, such as whether a tumor is aggressive or not. However, a new study published by a team from Tongji University, Tsinghua University, and Fudan University takes a different approach. Instead of asking what a tumor is, they asked how the tumor's abnormalities are arranged over time.

The researchers analyzed MRI scans from 246 patients who had been treated for either glioblastoma, the most aggressive form of the disease, or lower-grade glioma. They focused on eight specific measurements derived from the MRI images, such as how round or irregular the tumor's shape was, and how the brightness and texture varied across different scales. To make sense of this data, they used a statistical tool called SuStaIn, which stands for Subtype and Stage Inference. Imagine a timeline where different problems appear in a specific order. This tool does not just group patients into static boxes; it tries to figure out the sequence in which different abnormalities show up. It asks whether one group of patients tends to develop a specific shape irregularity first, followed by texture changes, while another group develops texture changes first, followed by shape issues. By running the data through this system many times to ensure the results were not a fluke, the researchers discovered that the patients naturally fell into two distinct groups based on the order of these imaging events.

The first group, which the researchers called S1, showed a pattern where the tumor's shape became irregular early on, followed by changes in the texture of the image at various scales. The second group, S2, displayed a different sequence. In these patients, the tumor became more compact and dense early in the process, accompanied by specific changes in the intensity and texture of the image. These two groups were not just different in their imaging patterns; they were different in their biology and their fate. Patients in the S2 group, those with the compactness-first pattern, had a significantly higher risk of death compared to those in the S1 group. Even when the researchers accounted for the patient's age, the grade of the tumor, and the presence of specific genetic mutations, the S2 pattern remained a strong predictor of a poorer outcome. The risk was not just a simple switch between two categories; the more likely a patient was to belong to the S2 group, the higher their risk of mortality became, creating a smooth gradient of danger.

To understand why these two groups behaved so differently, the team looked at the genetic activity inside the tumors of 194 patients who had matching tissue samples. They found that the tumors in the high-risk S2 group were driven by biological programs related to rapid growth and division. These tumors were busy making the machinery needed to build new cells, replicating their DNA, and separating their chromosomes. In contrast, the lower-risk S1 group showed genetic activity related to the brain's normal function. Their tumors were more connected to the networks that neurons use to communicate, involving signals that help brain cells talk to one another. This suggests that the S2 tumors are aggressively focused on multiplying, while the S1 tumors retain more of the characteristics of the brain tissue they grew from. The study also looked at the immune system's presence within the tumors but found that the differences between the two groups were small and not the main driver of their distinct behaviors.

The findings offer a new way to look at brain tumors, moving beyond static labels to see them as dynamic processes with different internal rhythms. The researchers emphasize that their work is a starting point for understanding these hidden patterns rather than a finished tool for clinical use. Because the study relied on a single set of historical data, the specific order of events they found is relative to that group and needs to be tested in other populations with different scanners and protocols. Furthermore, the study used existing data where the genetic and imaging information was already collected, so the results are a strong suggestion of a biological reality rather than a final proof of cause and effect. However, the connection between the shape of the tumor on an MRI and the molecular machinery inside it provides a compelling reason to keep exploring. By listening to the sequence of events that a tumor tells through an image, doctors may one day be able to predict a patient's future with greater accuracy and tailor treatments to the specific biological story of their disease.