Pathology-led digital collagen quantification for risk profiling of mesenteric disease in small-intestinal neuroendocrine tumors

Mesenteric disease is a major determinant of morbidity and operative complexity in small-intestinal neuroendocrine tumors (SI-NETs), yet routine pathology alone can be insufficient for risk profiling of mesenteric involvement and distinguishing mesenteric tumor deposits (MTDs) from nodal metastases. This study evaluated a pathology-led framework integrating standardized clinicopathologic reassessment with digital collagen quantification on picrosirius red (PSR) whole-slide images and an interface-based assessment of vessel-wall modification (VWM). This single-center retrospective cohort included 70 resected, well-differentiated SI-NET cases from 2010 to 2024. Mesenteric disease was present in 49 of 70 SI-NET cases (primary tumor with mesenteric metastasis, either nodal metastasis or mesenteric tumor deposit), whereas 21 of 70 were primary-only cases with no mesenteric metastasis. PSR collagen proportionate area (CPA) was quantified using region-of-interest annotation and color deconvolution for compartment-specific primary bowel-wall regions and for mesenteric lesions. Mesenteric lesions were typed on H&E using prespecified architecture-based criteria as MTDs versus nodal metastases. A binary patient-level analysis compared primary-only cases versus any mesenteric lesion, and a lesion-level analysis assessed discrimination of MTDs versus nodes using CPA and VWM. In the binary analysis, patients with mesenteric lesions had larger primary tumors (median 2.0 vs 1.5 cm; p = 0.0378; ROC AUC = 0.66) and markedly higher frequencies of primary vascular invasion (OR = 10.2; p = 0.0014) and perineural invasion (OR = 8.5; p = 0.00042). In contrast, primary-tumor CPA showed minimal discrimination for mesenteric involvement (ROC AUC 0.49 to 0.54). In the lesion-level analysis (19 mesenteric tumor deposits; 29 mesenteric nodal metastases), CPA was higher in MTDs than in nodes (median 21.1% vs 15.7%; p = 0.00158; ROC AUC = 0.709); a Youden-derived CPA threshold of at least 18% yielded sensitivity 0.77 and specificity 0.61. VWM was strongly enriched in MTDs (89.5% vs 41.4%; OR = 12.04; p = 0.000828), with sensitivity 0.895 and specificity 0.586, and concordant positivity for CPA ≥18% together with VWM most strongly supported an MTD diagnosis in this dataset. In this cohort, larger primary-tumor size and the presence of vascular and/or perineural invasion were associated with mesenteric involvement, whereas an interface-focused combination of VWM and higher lesion CPA supported distinction of MTDs from nodal metastases.