Ultrasound-Based Predictive Model to Assess the Risk of Orbital Malignancies.

OBJECTIVE: Ultrasound (US) is widely used for evaluating various orbital conditions. However, accurately diagnosing malignant orbital masses using US remains challenging. We aimed to develop an ultrasonic feature-based model to predict the presence of malignant tumors in the orbit.

METHODS: A total of 510 patients with orbital masses were enrolled between January 2017 and April 2023. They were divided into a development cohort and a validation cohort. In the development cohort (n = 408), the ultrasonic and clinical features with differential values were identified. Based on these features, a predictive model and nomogram were constructed. The diagnostic performance of the model was compared with that of MRI or observers, and further validated in the validation cohort (n = 102).

RESULTS: The involvement of more than two quadrants, irregular shape, extremely low echo of the solid part, presence of echogenic foci, cast-like appearance, and two demographic characteristics (age and sex) were identified as independent features related to malignant tumors of the orbit. The predictive model constructed based on these features exhibited better performance in identifying malignant tumors compared to MRI (AUC = 0.78 [95% CI: 0.73, 0.82] vs. 0.69 [95% CI: 0.64, 0.74], p = 0.03) and observers (AUC = 0.93 [95% CI: 0.90, 0.95] vs. Observer 1, AUC = 0.80 [95% CI: 0.76, 0.84], p < 0.01; vs. Observer 2, AUC = 0.71 [95% CI: 0.66, 0.76], p < 0.01). In the validation cohort, the predictive model achieved an AUC of 0.88 (95% CI: 0.81, 0.94).

CONCLUSION: The ultrasonic-clinical feature-based predictive model can accurately identify malignant orbital tumors, offering a convenient approach in clinical practice.