三阴性乳腺癌癌巢的三维重建与空间表型的解析OA

Objective:To build a three-dimensional(3D)reconstruction and quantification pipeline for cancer nests in triple-negative breast cancer(TNBC)from serial histological sections and characterize their spatial clustering patterns and morphological spectrum.Methods:Five surgical pathology specimens of TNBC were included.These specimens underwent serial sectioning,hematoxylin and eosin staining,and whole-slide imaging digitization.Cellular oriented data analysis was employed for multi-scale registration,and DeepLabv3+was uti-lized for semantic segmentation.A voxel stack was constructed based on pixel spacing,and"discrete cancer nests"were de-fined using 26-neighborhood connectivity to obtain scale-comparable 3D connected volumes.Within these statistical units,morphological parameters including volume,sphericity(ψ),and surface area-to-volume ratio(SA/V)were quantified.Spa-tial deviation of centroids across different scales was assessed using the 3D Ripley's L(r)-r function,and sample-level"morphology-scale"profiles were summarized using a robust z-score heatmap.Results:The two-dimensional(2D)cancer nest count was significantly higher than the 3D"independent count"(197,113 vs 27,427,ratio ≈ 7.2∶1),which indi-cates that cancer nests are spatially independent in 3D space.L(r)-r analysis showed that the centroids of cancer nests shif-ted from discrete distribution to clustered aggregation at approximately 10 μm;the values remained positive within 25～30 μm,indicating that short-and medium-range clustering constitutes a robust feature with interindividual variation in cluste-ring intensity.Analysis of merged data showed ψ values primarily distributed between 0.60～0.80(median ≈ 0.65),while SA/V were concentrated in the 0.60～1.00 μm-1 range(median ≈ 0.70 μm-1).At comparable volumes,ψ and SA/V exhibited a negative correlation,forming a continuous morphological spectrum from"more spherical with minimal sur-face irregularity"to"highly irregular with complex surface contours"(including laminar organization,annular structures with central lumina,bridging with fenestrations,and reticulated porous patterns).Sample-level heat maps revealed distinct quantity-driven and volume-driven profiles,along with a tail subpopulation characterized by large nests with high SA/V rati-os.Conclusion:The 3D-connected volume-defined independent cancer nests,verified by the 2D-to-3D cancer nest ratio,confirm the independence of 3D cancer nests,and reveal short-range clustering starting at approximately 10 μm in TNBC as well as a geometric morphological spectrum(including laminar organization,annular structures with central lumina,bridging with fenestrations,and reticulated porous patterns).By incorporating 3D independent counting,clustering initiation and in-tensity,ψ,and surface area-to-volume ratios,we classified samples into volume-driven,quantity-driven,and mixed sub-types,collectively constructing an interpretable and reusable 3D quantitative framework.