🤖 AI Summary
Accurate automatic segmentation of the pituitary gland and pituitary adenomas in MRI remains challenging due to their small anatomical scale, low contrast, and high inter-subject variability—hindering clinical deployment. Method: This systematic review critically evaluates deep learning (e.g., U-Net variants) and semi-automatic segmentation approaches, benchmarking performance using standard metrics (e.g., Dice coefficient) and analyzing impacts of field strength, patient age, and tumor size. Contribution/Results: State-of-the-art automatic methods achieve up to 96.41% Dice for adenoma segmentation but exhibit poor robustness on normal pituitary tissue; semi-automatic methods demonstrate superior stability. Crucially, this work identifies systemic gaps in data reporting completeness, annotation standardization, and cross-center generalizability—previously unreported. It proposes a foundational framework: a high-quality, multi-center, multi-sequence, age-stratified benchmark dataset to enable development and clinical translation of accurate, robust, and deployable pituitary segmentation models.
📝 Abstract
Purpose: Accurate segmentation of both the pituitary gland and adenomas from magnetic resonance imaging (MRI) is essential for diagnosis and treatment of pituitary adenomas. This systematic review evaluates automatic segmentation methods for improving the accuracy and efficiency of MRI-based segmentation of pituitary adenomas and the gland itself. Methods: We reviewed 34 studies that employed automatic and semi-automatic segmentation methods. We extracted and synthesized data on segmentation techniques and performance metrics (such as Dice overlap scores). Results: The majority of reviewed studies utilized deep learning approaches, with U-Net-based models being the most prevalent. Automatic methods yielded Dice scores of 0.19--89.00% for pituitary gland and 4.60--96.41% for adenoma segmentation. Semi-automatic methods reported 80.00--92.10% for pituitary gland and 75.90--88.36% for adenoma segmentation. Conclusion: Most studies did not report important metrics such as MR field strength, age and adenoma size. Automated segmentation techniques such as U-Net-based models show promise, especially for adenoma segmentation, but further improvements are needed to achieve consistently good performance in small structures like the normal pituitary gland. Continued innovation and larger, diverse datasets are likely critical to enhancing clinical applicability.