Habitat Imaging Combined with Multiscale Peritumoral Radiomics and Pathological STAS Status for Predicting Early Postoperative Recurrence in Stage T1 Non-Small Cell Lung Cancer

Lung cancer remains the leading cause of cancer-related death worldwide, and for patients with early-stage disease, surgery is often the standard cure. However, even after a successful operation, some patients face a significant risk that the cancer will return. One specific biological feature, known as spread through air spaces, acts as a powerful warning sign. This occurs when cancer cells detach from the main tumor and float freely into the tiny air sacs of the lung, a behavior that makes the disease more likely to come back. The problem is that doctors can only confirm this dangerous feature after the tumor has been removed and examined under a microscope, leaving them without a way to predict the risk before the surgery begins. To solve this, researchers have turned to a field called radiomics, which treats medical images not just as pictures, but as vast sources of hidden data. By analyzing the subtle patterns of light and texture within a scan, computers can detect details invisible to the human eye, offering a potential window into the tumor's true nature before a single incision is made.

A team of researchers at Shanxi Bethune Hospital in China has developed a new method to predict whether early-stage lung cancer will return after surgery. They focused on patients with stage T1 non-small cell lung cancer, a group that typically has a good prognosis but still faces a non-negligible risk of recurrence. The team created a sophisticated computer model that combines three distinct types of information: the patient's clinical details, the biological status of the spread through air spaces, and a deep analysis of the tumor's internal structure and its immediate surroundings. Their goal was to build a tool that could help doctors identify high-risk patients earlier, allowing for more personalized follow-up plans and treatment decisions.

To build this model, the researchers looked back at the records of 216 patients who had undergone surgery for early-stage lung cancer between 2019 and 2022. They divided these patients into two groups: a larger group to train the computer and a smaller group to test how well the computer learned. The team started by examining the patients' chest CT scans, which are detailed 3D images of the lungs. Instead of treating the tumor as a single, uniform blob, they used a technique called habitat imaging. This method breaks the tumor down into smaller, distinct sub-regions based on how the pixels look and behave, much like a geologist might map different types of soil within a single field. By grouping these tiny areas together, the computer could see the complex, uneven landscape inside the tumor that a standard scan would miss.

The researchers also looked beyond the tumor itself. They analyzed the tissue immediately surrounding the cancer, creating rings of analysis at distances of one, three, and five millimeters from the tumor edge. They knew that the environment right next to a tumor often holds clues about how aggressively it is growing. They extracted thousands of numerical features from these zones, measuring everything from the shape of the tumor to the texture of the tissue. They then combined these imaging details with the patients' medical history, such as age and gender, and the results of the pathological examination that confirmed whether the spread through air spaces was present.

The team tested several different computer learning algorithms to see which one could best predict recurrence. They found that a model combining all these elements—clinical factors, the internal "habitats" of the tumor, and the features from the three-millimeter ring around the tumor—performed the best. In their testing group, this combined model correctly identified the risk of recurrence with an accuracy rate that translated to a score of 0.865. This was significantly better than models that relied on just the clinical data, just the internal tumor patterns, or just the surrounding tissue alone. The study showed that the three-millimeter zone around the tumor provided the most useful information about the surrounding environment, suggesting that the immediate neighborhood of the tumor is critical for understanding its behavior.

The researchers also confirmed that certain factors were independent predictors of recurrence. These included the presence of a specific sign on the scan called a vacuole sign, the patient's gender, the presence of spread through air spaces, the way the tumor pulled on the nearby lung lining, and the maximum size of the tumor. When the computer model integrated the spread through air spaces status with the detailed imaging data, it created a much clearer picture of the risk than any single piece of information could provide on its own. The model was able to distinguish between patients who would remain cancer-free and those who would experience a return of the disease with a high degree of reliability.

While the results are promising, the authors note that their study has limitations. It was conducted at a single hospital with a moderate number of patients, and the data came from a specific population, which means the model needs to be tested on larger and more diverse groups of people before it can be used widely. Additionally, the process of mapping the tumor regions was done manually, which is time-consuming, though future advances in automated technology may speed this up. Despite these hurdles, the study demonstrates that combining different layers of information—from the microscopic biology of the tumor to the macroscopic patterns in the scan—creates a powerful tool for prediction.

This work offers a practical path forward for managing early-stage lung cancer. By using a non-invasive approach that analyzes the tumor's internal diversity and its immediate environment, doctors may soon be able to predict recurrence risk with greater precision. This could help tailor post-surgery care, ensuring that patients who are at higher risk receive the necessary attention and treatment, while sparing those at lower risk from unnecessary interventions. The study suggests that the future of cancer care lies not just in looking at the tumor, but in understanding the complex story it tells through its shape, its surroundings, and its hidden internal structure.