Assessment of intratumoral collagen in pituitary neuroendocrine tumors (pituitary adenomas/PitNets) via digital pathology and its correlation with intraoperative consistency
摘要
The success of endoscopic transsphenoidal resection of pituitary neuroendocrine tumors (pituitary adenomas/PitNets) depends largely on tumor consistency, as fibrous tumors pose greater surgical difficulty and are associated with higher morbidity. However, histological correlates of intraoperative consistency remain inconsistently defined. This study evaluated the relationship between intraoperative tumor consistency and collagen content using digital image analysis.
MethodsBiopsy samples from patients undergoing surgery for pituitary adenomas/PitNets were included. Tumor consistency was assessed intraoperatively by the same neurosurgeon and classified as soft, firm, or fibrous. Histological sections were stained with Masson’s trichrome, and digital image analysis was performed using QuPath software. A trained pixel classifier artificial neural network differentiated collagen, tumor tissue, and background. Collagen content, as a surrogate marker of fibrosis, was quantified as a percentage of total tumor area. All analyses were conducted blinded to intraoperative assessments.
ResultsA total of 69 tumors were analyzed: 30 soft (43.5%), 22 firm (31.9%), and 17 fibrous (24.6%). Median collagen content increased progressively across groups (6.3%, 8.3%, and 22.0%, respectively). Fibrous tumors demonstrated significantly higher collagen content compared with soft and firm tumors (p < 0.05), while no significant difference was observed between soft and firm tumors. Collagen content was not associated with tumor subtype or patient age, and its deposition was heterogeneous between paired tissue blocks (CV = 59.83%).
ConclusionDigital collagen quantification provides an objective histological correlate of intraoperative tumor consistency, validating the distinction between soft and fibrous tumors, and establishing a reference standard for future studies aimed at developing preoperative predictive models.