Enhanced prediction of postlaminar optic nerve invasion in retinoblastoma using high-resolution MRI: integrating qualitative imaging features, optic nerve measurements and tumor radiomics.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
OBJECTIVES: This study aimed to develop a model for predicting postlaminar optic nerve invasion (PLONI) in retinoblastoma patients using MR imaging. The model was designed to incorporate recent advanced diagnostic techniques, including tumor radiomics, manual and automatic distal optic nerve width measurements, and qualitative PLONI assessments. MATERIALS AND METHODS: This retrospective, multicenter case-control study included patients with histopathologically confirmed PLONI and high-spatial-resolution pretreatment MR images. PLONI appearance was qualitatively assessed by two expert radiologists. Quantitative measurements of optic nerve diameter were automatically obtained using a segmentation model and compared to manual measurements by the radiologist using the area under the curve (AUC). The best-performing optic nerve measurements, along with clinical features and radiomic features derived from tumor delineations, were used for prediction modeling. RESULTS: One hundred twenty-eight retinoblastoma patients (median age, 22 months [range, 0-113], 58 female) were included in this study. This resulted in 30 eyes with histopathologically proven PLONI and 146 control eyes without PLONI. Manual distal optic nerve diameter measurements were chosen for prediction modeling, as this significantly outperformed the automatic approach for detecting PLONI, achieving an AUC of 0.85 compared to 0.68 (p < 0.001) for the automatic method. The final logistic regression model, which integrated radiomic features and clinical features including optic nerve width and PLONI appearance MR imaging, achieved an AUC of 0.90 (95% CI: 0.86, 0.95, SD 0.07), outperforming the other models. CONCLUSION: This study developed a high-performing model for predicting PLONI, highlighting the value of advanced MR imaging and the integration of both qualitative and quantitative features. By enhancing risk assessment, this model could provide clinicians with an additional tool for guiding treatment decisions, optimizing early intervention strategies, and ultimately improving patient care in retinoblastoma management. KEY POINTS: Question Accurately detecting postlaminar optic nerve invasion in retinoblastoma remains challenging. Improved imaging-based prediction methods are needed to optimize risk assessment and potentially treatment decisions. Findings Manual distal optic nerve width measurements combined with qualitative imaging features and radiomics achieved high diagnostic performance for predicting postlaminar optic nerve invasion in retinoblastoma. Clinical relevance This study improves the prediction of postlaminar optic nerve invasion in retinoblastoma by integrating qualitative imaging assessments, optic nerve measurements, and radiomics, enhancing early detection and thus potentially reducing unnecessary aggressive treatments.