MRI deep learning radiomics and pathologic lymphocytic reaction predict 3-year distant metastasis in rectal cancer

Fusion model: MRI deep learning radiomics + lymphocytic reaction predicted 3-yr distant metastasis in rectal cancer (AUC 0.911/0.884, n=190). LR an independent predictor; DLR inversely correlated (r=-0.44).

• The fusion model integrating lymphocytic reaction and deep learning radiomic signatures achieved AUCs of 0.911 (training) and 0.884 (validation) for predicting 3-year distant metastasis-free survival.
• Lymphocytic reaction was confirmed as an independent predictor of 3-year distant metastasis-free survival (p<0.05).
• A moderate inverse correlation (r=-0.44, p<0.001) between deep learning radiomic signatures and lymphocytic reaction suggests an immune-related background for the imaging biomarkers.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.