Abstract PS2-10-09: Evaluation of spatial collagen morphometry in TNBC versus non-TNBC biopsies: a cross-sectional SHG/TPE and AI-based analysis of tumor and non-tumor regions

Abstract BACKGROUND: Triple-Negative Breast Cancer (TNBC) is a heterogeneous and aggressive subtype of breast cancer characterized by the absence of estrogen receptor, progesterone receptor, and HER2 amplification. The tumor microenvironment in breast cancer plays a critical role in disease progression Acerbi et al. Integr Biol (Camb). 2015 Oct;7(10):1120-34, yet the specific spatial characteristics and morphometric patterns of collagen within different tissue regions remain poorly defined, especially when comparing TNBC and non-TNBC pathologies. Quantitative collagen analysis using second harmonic generation/two-photon excitation (SHG/TPE) imaging and AI-based collagen morphometric evaluation provides an opportunity to characterize fibrosis patterns beyond conventional histopathology. This study aims to delineate and compare collagen morphometric features between tumor and non-tumor regions in TNBC biopsies versus those in non-TNBC biopsies, and to describe region-specific patterns of fibrosis and their similarities and differences. METHODS: 266 unstained breast cancer biopsy samples acquired from Dept of Surgery, National Taiwan University Hospital were analysed using SHG/TPE microscopy and AI analysis. Samples were divided into TNBC (n=21) and non-TNBC (n=245) groups. Collagen morphometric features were identified using a proprietary AI-based analysis using the Genesis®200 scanner (HistoIndex Pte Ltd, Singapore). Four anatomically distinct tissue regions were evaluated: tumor, lobule and duct, stroma-fat, and stroma-fibrosis, with the latter three comprising the non-tumor regions. Normalised Median Relative Differences in feature values were computed to compare collagen parameters between TNBC and non-TNBC groups for each region (Parameter-wise normalization scaled median differences within the scale of -1 to +1). The Wilcoxon rank-sum test was used to assess statistical significance. RESULTS: Collagen morphometric analysis revealed region-specific differences between TNBC and non-TNBC biopsies. In tumor regions, collagen parameters were largely comparable between the two groups, with no significant differences detected. In the non-tumor stroma-fibrosis region, some parameters related to collagen fibre distribution showed numerical differences favouring the TNBC group, but these differences did not reach statistical significance. In contrast, the lobule and duct region demonstrated statistically significant differences in multiple collagen morphometric features. Specifically, TNBC biopsies exhibited increased levels of collagen strings (defined as single collagen-connected areas comprising one or more intersecting fibres) and string aggregation features, including those involving various string dimensions such as thick and thin strings. Additionally, several parameters reflecting collagen fibre distribution were significantly elevated in TNBC samples compared to non-TNBC. CONCLUSIONS: SHG/TPE-based AI analysis demonstrates that TNBC is associated with a distinct fibrosis profile, particularly characterized by significant alterations in collagen morphometric features within the lobule and duct regions of the non-tumor compartment compared to non-TNBC cases. These region-specific differences in collagen architecture likely reflect underlying desmoplastic remodelling unique to TNBC and may serve as candidate imaging biomarkers for subtype differentiation. Further exploration of these spatial collagen morphometric patterns may enhance understanding of tumor-stroma dynamics and support the development of targeted therapeutic strategies in breast cancer. Citation Format: K. Akbary, L. Jiaojiao, R. Yayun, D. Tai, C. Hsiao, K. Huang. Evaluation of spatial collagen morphometry in TNBC versus non-TNBC biopsies: a cross-sectional SHG/TPE and AI-based analysis of tumor and non-tumor regions abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS2-10-09.