To investigate the efficacy of habitat-based histogram features derived from APTw MRI for predicting BRAF mutation status in rectal cancer and its potential for prognostic stratification.
Approach:
Study Design: Prospective enrollment of 269 rectal cancer patients, divided into training (n=188) and testing (n=81) sets, categorized by BRAF mutation status.
Data Analysis: K-means clustering was used to partition tumors into sub-regions, and histogram features were extracted and filtered using stepwise regression. Predictive performance was evaluated using ROC curve analysis and DCA.
Survival Analysis: Kaplan-Meier survival curves assessed 2-year disease-free survival based on a combined nomogram.
Key Findings:
The K-means model with K=7 demonstrated optimal predictive power for BRAF mutation status, outperforming models with K=5 or K=6.
The nomogram integrating APTw/ADC features and clinical factors achieved an AUC of 0.83 (95% CI: 0.707-0.952), significantly higher than the habitat-based APTw model (0.733), habitat-based ADC model (0.71), or clinical factors model (0.661) (all P < 0.05).
Patients in the mutant group had significantly worse 2-year disease-free survival (DFS) compared to the wild-type group (P < 0.05).
Interpretation:
Habitat-based APT/ADC histogram features combined with clinical factors predict BRAF mutation status and disease-free survival in rectal cancer patients.
Limitations:
The study was limited to a single institution.
The sample size for the mutant group was relatively small.
Conclusion:
The study demonstrates the potential of habitat-based APT/ADC histogram features for non-invasive prediction of BRAF mutation status and prognostic stratification in rectal cancer.
