What specific radiomic features does the new MRI model use to stratify risk?

The new MRI-based model for risk stratification in locally advanced rectal cancer (LARC) combines two types of radiomic features: habitat heterogeneity features from within the tumor and peritumoral features from the tissue surrounding the tumor. This approach aims to capture the tumor microenvironment better than single-feature models.

What the research says

A multicenter retrospective study of 290 LARC patients developed a nomogram that integrates MRI-based habitat heterogeneity and peritumoral radiomics ⁵. Habitat heterogeneity was quantified by applying K-means clustering (with k=3) to intratumoral radiomic features, which groups different regions within the tumor based on their imaging characteristics ⁵. Peritumoral radiomic features were extracted from regions at 1, 2, and 3 mm margins around the tumor volume ⁵. These features were then filtered and reduced using LASSO regression, and logistic regression and support vector machine classifiers were used to build intratumoral, peritumoral, and habitat models ⁵. The best-performing model was combined with clinical variables to construct the final nomogram ⁵. Other studies have also used radiomic features from T2-weighted MRI and diffusion-weighted imaging to predict treatment response, but the specific model described here uniquely combines habitat and peritumoral features ⁹¹⁰¹¹.

What to ask your doctor

• What are the specific radiomic features used in the MRI model for my risk stratification?
• How does the model incorporate intratumoral habitat heterogeneity and peritumoral features?
• Is this MRI model available at my treatment center, and how might it affect my treatment plan?
• What is the evidence that this model improves risk stratification compared to standard MRI staging?
• How reliable are the predictions from this model for guiding decisions about total neoadjuvant therapy?