To evaluate the diagnostic performance of deep learning−based radiomics (DL) and hand−crafted radiomics (HCR) in differentiating benign from malignant orbital tumors, highlighting its potential impact on clinical practice.
Key Findings:
Homogeneous enhancement and ill-defined/infiltrative margins were identified as independent CT features differentiating benign from malignant tumors.
The HCR, DL, fused, and nomogram models achieved AUCs of (0.859/0.816), (0.957/0.826), (0.986/0.811), and (0.975/0.837) in training and test cohorts, respectively.
The fused model outperformed single radiomics approaches in accuracy, and the nomogram demonstrated superior predictive performance, with higher clinical utility as indicated by DCA.
Interpretation:
The integration of deep learning and traditional radiomics enhances the ability to non-invasively differentiate between benign and malignant orbital tumors, potentially improving clinical decision-making by providing reliable diagnostic alternatives.
Limitations:
The study was limited to a single institution, which may affect the generalizability of the findings.
The retrospective nature of the study may introduce selection bias, potentially impacting the robustness of the conclusions drawn.
Conclusion:
The study supports the use of a fused model and nomogram for accurate, non-invasive differentiation of orbital tumors, aiding in clinical decision-making.
