Therapy-induced tumor regression heterogeneity for early prediction of response and prognosis in HER2-positive breast cancer.

BACKGROUND: Early prediction of pathologic complete response (pCR) during neoadjuvant chemotherapy (NAC) in HER2-positive breast cancer remains challenging using conventional whole-tumor imaging approaches, particularly when substantial tumor shrinkage occurs on mid-NAC MRI (acquired after two cycles of NAC).

METHODS: This retrospective study included 163 patients with HER2-positive breast cancer who underwent dynamic contrast-enhanced MRI at pre-NAC and mid-NAC. A therapy-induced tumor regression region (TRR) was defined as the regressed tumor volume between pre- and mid-NAC MRI. Tumor regression heterogeneity (TRH) was quantified within this region using a habitat-based radiomic approach. Logistic regression models were constructed and evaluated in training (n = 115) and validation (n = 48) cohorts.

RESULTS: pCR was achieved in 71 patients (43.6%). The TRH model demonstrated superior predictive performance compared with pre-NAC, mid-NAC, tumor regression region-based, and delta models, achieving an AUC of 0.83 (95% CI: 0.70-0.97) in the validation cohort. An integrated nomogram combining TRH with clinicopathologic factors further improved predictive performance.

CONCLUSIONS: Quantification of spatial heterogeneity within therapy-induced tumor regression regions on mid-NAC MRI may enable improved prediction of treatment response during neoadjuvant therapy.