Novotny, Brenna C.

Abstract Background Multiplexed immunofluorescence (MxIF) enables high-dimensional immune cell phenotyping and detailed characterization of the tumor immune microenvironment (TIME), but complex cyclic staining can introduce signal degradation and uncertainty in cell-level measurements. In routine practice, pathologists rely on hematoxylin and eosin (H&E) staining as the primary reference for tissue morphology and frequently cross-reference H&E to interpret MxIF findings. Methods We present a framework for aligning H&E and MxIF images at the tissue microarray (TMA) core level to support cross-modal analysis. Cell nucleus detections from each modality are used as anchor points, formulating the task as a point-set alignment problem. Coherent Point Drift (CPD) is applied for global alignment, followed by graph-matching–based refinement to improve local consistency. Evaluation on ovarian TMAs demonstrates consistent alignment performance for both restained and serial sections. We further explore the use of restained H&E as a reference for MxIF interpretation and investigate virtual H&E generation from MxIF data as a complementary approach when restained H&E is unavailable. Results The aligned images enable quantitative assessment of cross-modal feature concordance between MxIF-derived measurements and H&E morphology. The generated virtual H&E demonstrated similar cell population compared to the real H&E. Conclusions Cell-centric alignment can facilitate integrative multimodal histopathology analysis for TMA cores in spatial co-localization, cell feature validation, and virtual H&E generation.

Abstract Deep learning models enable the prediction of clinical endpoints from whole-slide images (WSIs), but many such models function as “black boxes”, lacking transparency about whether and which histomorphological patterns drive their predictions, hindering interpretability and clinical adoption. Here we propose a human-in-the-loop explanation framework, MorphoXAI, which provides both local and global interpretability for deep learning models by incorporating human-expert interpretations. At the global level, it reveals the histomorphological patterns on which the model consistently relies to distinguish between classes of WSIs, as well as the patterns associated with confusion between classes. At the local level, it indicates which of these patterns are used in the prediction of an individual WSI and which regions within the slide correspond to such patterns. We validated our method on a deep learning model trained for ovarian tumor histologic subtype prediction. The results show that our framework generates explanations that accurately reflect the histomorphology underlying the model’s predictions at both global and local levels. For interpretability and clinical utility in diagnostic contexts, human evaluation results showed that our explanations were easy to interpret, rich in diagnostic features, and directly helpful for diagnostic decision-making, thereby enhancing pathologist-AI collaboration. Our work highlights that unifying global and local explanations and grounding them in expert-interpreted morphology enhances the interpretability and verifiability of deep learning models, thereby facilitating the transparent deployment of such models in clinical practice.

Abstract Background: The tumor microenvironment (TME) plays a pivotal role in ovarian cancer (OC) progression and prognosis, while also contributing to biological differences between primary and metastatic tumors. This study examines OC TMEs to uncover associations with clinical prognosis and tumor biology by analyzing paired primary and metastatic tumors from the same patients. Methods: This study focused on 57 stage-IIIC/IV OC patients treated at the Mayo Clinic, all of whom underwent surgical debulking and adjuvant therapies. All patients provided informed consent, and the study received IRB approval. A tissue microarray (TMA) was specifically designed and constructed, containing paired primary and metastatic tissue cores from the same patients. 147 tissue cores from the constructed TMA were profiled using a 35-antibody panel, with 17 antibodies for cell phenotyping and 18 for functional state characterization. Profiling was performed using Cell DIVE, a multiplexed imaging solution at the single-cell level developed by Leica Microsystems. After cell segmentations and classifications, detailed spatial niche and regional analyses were performed at the single-cell level. Immune cell populations, as well as their TME distributions, were characterized to conduct prognostic associations and contrast primary and metastasis differences. Results: A total of 1,292,988 total cells (after excluding low-density signals and ambiguous classifications) were analyzed, with a median of 8,033 cells per core. Using spatial neighborhood analysis, 12 distinct spatial niches were identified with specific cell types and/or unique functional states. Among these, a macrophage- and fibroblast-enriched niche at the tumor-stroma interface was associated with improved overall survival (p=0.008). Another rare Treg-enriched niche, along with other immune cells, demonstrated protective effects, particularly in patients with residual macroscopic tumors after debulking surgery (p=0.05), while no significant effects were observed in those without visible residual disease (p=0.8). When comparing primary and metastatic tumors, metastatic lesions exhibited significantly increased tumor mass (p=2e-4) and a notable reduction in helper T cells (p=0.04) in tumor areas, as well as significantly increased fibroblast cells in tumor-stroma interfacing regions (p=0.002), reflecting distinct immunosuppressive and tumor-promoting shifts. Conclusion: This study reveals spatial and cellular differences in primary and metastatic OC TMEs, distribution differences linked to prognosis and surgical outcomes. These findings emphasize the potential of spatially resolved TME profiling to refine adjuvant and neoadjuvant therapeutic strategies. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6266.

Abstract Purpose: Tumor-infiltrating lymphocytes are associated with a decreased risk of recurrence in human papillomavirus–associated oropharyngeal squamous cell carcinoma. The composition and spatial distribution of tumor-infiltrating lymphocytes and tumor-infiltrating immune cells are not well characterized. Experimental Design: Formalin-fixed, paraffin-embedded primary and lymph node (LN) tumor tissues from 10 progressors (cases) and 10 matched nonprogressors (controls) were interrogated by imaging mass cytometry. Immune, stromal, and tumor cells were quantified from selected regions of interest using machine learning. Nearest neighbors, cell–cell interactions, and niche analyses were performed. Results: In primary regions of interest, immune cell, lymphocyte, T cell, CD8+ T cell, and innate cell prevalence was significantly greater in controls. High prevalence of immune cells, lymphocytes, innate cells, and CD4+ T cells in primary tissues was significantly associated with increased time to event (TTE). Although primary and LN prevalence of T cells, CD4+ T cells, CD8+ T cells, macrophages, and tumor cells were significantly correlated, differences in LNs were neither significant nor associated with TTE. Average distances between T cells and the nearest B cells and between lymphocytes and the nearest tumor cells were decreased in control primary tissues. Interactions between B and T cells were less organized in primary tissues from cases. A niche predominantly comprising lymphocytes was associated with longer TTE. Conclusions: In human papillomavirus–associated oropharyngeal squamous cell carcinoma, immune cell subset prevalence in primary tumors differs with outcome and is associated with TTE. Interactions between B cells and both T-cell subsets are associated with longer TTE, underscoring the importance of active intratumoral immune responses in outcomes.

Abstract Motivation Multiplex immunofluorescence (MxIF) enables the quantification of multiple protein markers at a single-cell level while preserving spatial information, offering a powerful tool for studying tissue microenvironments. However, the flexibility in MxIF panel design poses challenges in standardizing cell phenotyping. Results We present SpaFlow, an efficient, customizable pipeline for unsupervised clustering and classification of MxIF data, implemented using Nextflow. SpaFlow performs quality control, clustering, and postclustering analysis on segmented and quantified MxIF data, facilitating reproducible and scalable analyses across various computing platforms. The SpaFlow pipeline integrates three clustering and classification packages—Seurat, SCIMAP, and CELESTA—each providing unique methodologies for identifying cell types based on phenotypic markers. A novel “meta-clustering” approach condenses clusters across multiple regions of interest into common meta-clusters, streamlining the cell-type identification process in large datasets. SpaFlow’s robust quality control steps, including signal summation and cell density filtering, mitigate artifacts that may impact clustering accuracy. We demonstrate the utility of SpaFlow in a case study involving 297 ovarian tumor cores, where SpaFlow successfully identified biologically meaningful cell populations, including tumor-infiltrating lymphocytes, efficiently and rapidly. Additionally, SpaFlow’s reproducibility is validated using serial tonsil sections, confirming its capability to consistently identify distinctive cell populations across matched ROIs. Availability and implementation SpaFlow is freely available with detailed documentation and examples at https://github.com/dimi-lab/SpaFlow.

Abstract Multiplexed Immunofluorescence (MxIF) enables detailed immune cell phenotyping, providing critical insights into cell behavior within the tumor immune microenvironment (TIME). However, signal integrity can be compromised due to the complex cyclic staining processes inherent to MxIF. Hematoxylin and Eosin (H&E) staining, on the other hand, offers complementary information through its depiction of cell morphology and texture patterns and is often visually cross-referenced with MxIF in clinical settings. In this study, we proposed a novel framework to align H&E and MxIF images for precise cross-modal cell feature validation. Using cell detection outputs from each modality as anchors, we formulated the multimodal image registration problem as point set alignment. Coherent Point Drift (CPD) is employed for initial alignment, followed by Graph Matching (GM) for refinement. Evaluations on ovarian cancer tissue microarrays (TMAs) demonstrate that our method achieves high alignment accuracy, enabling reliable validation of cell-level features across modalities for both restained and serial sections. Our results indicate that restained H&E enhances confidence in findings derived from MxIF. Additionally, we demonstrated the feasibility of generating high-quality virtual H&E images from MxIF data when restained H&E is unavailable, offering a viable alternative for integrated multimodal analysis.