Emerging multicenter research suggests that tumor structure habitat (SH) phenotypes derived from radiomic patterns on breast MRI may enhance pre-treatment risk stratification beyond clinicopathological factors for women with stage 2 and stage 3 breast cancer.
For the retrospective multicenter study, which was recently published in European Radiology, researchers reviewed tumor SH phenotype data derived from pre-treatment breast MRI scans in 842 women (mean age of 50.7) with stage 2 or stage 3 breast cancer. Examining radiomic patterns in the tumor central and peripheral subregions, the study authors identified two tumor SH phenotypes (SH-phenotype-2 and SH-phenotype 3) and compared them to a baseline radiomics model for correlation with recurrence-free survival (RFS).
In the discovery cohort of 415 women, the study authors noted 43 cases of recurrence and eight non-recurrence deaths. For the external testing cohort, there were 41 recurrences and five non-recurrence deaths, according to the study.
In comparison to the baseline radiomics model, the researchers found in a multivariable assessment (which included tumor size, clinical subtype and nodal involvement) that the model incorporating SH-phenotype 2 was associated with a significant reduction in recurrence-free survival (RFS) (hazard ratio (HR) of 2.29) as was the model incorporating SH-phenotype 3 (HR of 3.45). The study authors also noted a 2.44 HR for cumulative incidence of recurrence (CIR) with SH-phenotype 2 and a 6.05 HR for CIR with SH-phenotype 3.
“In two independent and highly heterogeneous cohorts, SH-phenotypes with elevated risk were consistently associated with decreased RFS rates and increased CIR, providing independent prognostic value beyond the classical clinicopathological factors,” noted lead study author Guangsong Wang, MD, who is affiliated with the Department of Radiology at Xiang’an Hospital of Xiamen University and the School of Medicine at Xiamen University in Xiamen, China, and colleagues.
Three Key Takeaways
• SH-phenotyping adds independent prognostic value beyond standard clinicopathology. In two large, independent cohorts, breast MRI-derived tumor structural habitat phenotypes (SH-2 and SH-3) remained significantly associated with worse recurrence-free survival and higher cumulative incidence of recurrence even after adjusting for tumor size, clinical subtype, and nodal status—suggesting these radiomic patterns capture risk information that conventional staging misses.
• Risk elevation is substantial and phenotype-dependent. SH-phenotype 3 carried a notably higher risk than SH-phenotype 2 (HR ~3.45 vs. 2.29 for RFS; ~6.05 vs. 2.44 for CIR), indicating a graded risk stratification that could help identify patients who may benefit from more aggressive treatment or closer surveillance.
• The model doesn't yet generalize to triple-negative breast cancer. No significant association was found between SH-phenotypes and outcomes in TNBC patients, so this tool shouldn't be applied uniformly across subtypes. Larger TNBC-specific cohorts are needed before drawing conclusions for that population, according to the study authors.
While pointing out that the consistency of risk stratification withSH-phenotypes in subgroups including T-stage and pathological grade data, the researchers said there was no statistically significant association between outcomes and the SH-phenotypes in women with triple-negative breast cancer.
“This negative finding highlights the need for subtype-specific investigations and warrants further exploration in larger triple-negative breast cancer cohorts,” added Wang and colleagues.
(Editor’s note: For related content, see “Video: Stamatia Destounis, MD, FACR, Discusses Key Changes for the Updated BI-RADS System,” “Breast Imaging in Focus: Key Pearls from New Guidelines on Proliferative Breast Lesions with Atypia and LCIS” and “Breast MRI Study Reveals 29 Percent Reduction in Scan Time with Deep Learning Reconstruction and Multi-Shot DWI.”)
In regard to study limitations, the authors acknowledged the potential impact of a semiautomatic tumor segmentation method upon radiomic feature stability and a modest number of recurrence events preventing analysis of the impact of tumor structure habitat phenotypes across women of different races and ethnicities. The researchers also noted limitations of a single DCE-MRI scan for complete assessment of the biological leading edge and tumor core.
