Automated segmentation of pituitary adenomas, pituitary gland, and internal carotid arteries on routine coronal contrast-enhanced T1-weighted MRI: a single-sequence feasibility study - Summary - MDSpire
Advertisement
Automated segmentation of pituitary adenomas, pituitary gland, and internal carotid arteries on routine coronal contrast-enhanced T1-weighted MRI: a single-sequence feasibility study
To evaluate the feasibility and clinical relevance of Swin-Unet and nnU-Net for automated segmentation of pituitary adenomas, pituitary gland, and internal carotid arteries using contrast-enhanced T1-weighted coronal MRI.
Approach:
Performance Metrics: Model performance was assessed using the Dice similarity coefficient (DSC), 95th percentile Hausdorff distance (HD95), true-positive rate (TPR), and false-positive rate (FPR).
Key Findings:
Mean Dice similarity coefficient was 0.70 (95% CI: 0.66-0.74) for Swin-Unet and 0.69 (95% CI: 0.65-0.73) for nnU-Net.
Mean Hausdorff distance was 5.21 mm (95% CI: 4.40-6.03) for Swin-Unet and 5.30 mm (95% CI: 4.44-6.16) for nnU-Net.
Interpretation:
Automated segmentation of pituitary adenomas, pituitary gland, and internal carotid arteries using single-sequence T1CE coronal MRI is feasible, with moderate voxel-level accuracy and reliable slice-level recognition.
Limitations:
Segmentation accuracy was lower for the pituitary gland.
Performance may decline when applied to heterogeneous imaging data.
Conclusion:
The study indicates the potential utility of automated segmentation as an adjunct tool for radiologic assessment and surgical planning.
In a study of 50 echocardiography reports, GPT-5 mini extracted 55 cardiovascular fields from free-text echocardiography reports with 92.5% exact-match agreement with expert annotation.
