When the Pathologist Becomes an Architect: The Era of Spatial Pathology

Beyond Cells: Maps, Trajectories, and New Ways to Understand Tissue

We are witnessing a profound shift in how we interpret histology. Traditionally, the pathologist was a local observer—analyzing two-dimensional sections under the microscope, focusing on individual cells, and interpreting focal patterns. Today, that role is evolving.

Thanks to spatial pathology, we no longer just look at cells—we map them. We don't examine fragments in isolation; we contextualize each cell within a biological neighborhood, revealing how it interacts, communicates, and positions itself within the tissue. The classical principle that “form follows function” takes on a new dimension—now guided by spatial coordinates and molecular data layered over traditional histology (1).

As Sandro Santagata aptly put it: “H&E is still our comfort food... but now we add layers of molecular information.” That phrase says it all—we still value the foundation, but we enrich it with an architecture of data that was previously invisible (2).

From 2D Histology to 3D Models: A Shift in Perspective

Viewing tissue in three dimensions isn’t just a technical upgrade—it’s a paradigm shift. Traditional 2D histology, as valuable as it has been, offers only a partial snapshot. A tissue section is just that: a slice. And in that fragmentation, we often lose the true structure.

What appears on a slide as isolated tumor buds may, in fact, be hidden extensions of a much larger mass. What seems like scattered lymphoid follicles in 2D, when reconstructed in 3D, reveals itself as a continuous and functional immune network. That’s the core insight: in 2D we see events; in 3D we understand relationships (3).

Thanks to techniques like light-sheet microscopy, we can now reconstruct entire tumors, clearly visualizing how malignant and immune cells interact in their native environment (4). This volumetric imaging enables the detection of subtle but critical features—invaded vessels, compromised margins, microscopic nests—that might otherwise be missed.

But this isn’t just about seeing more—it’s about clinical relevance. Recent studies show that 3D models offer superior classification of tumor aggressiveness by capturing the full architecture of glands, fibers, and cellular distributions within the tumor microenvironment(3).

And that’s where the pathologist’s role evolves. We’re no longer readers of flat slides—we become navigators of tissue landscapes, exploring volumes where every structure has depth, context, and continuity. And that 3D perspective can make the difference between diagnostic uncertainty… and a confident, actionable clinical decision(3).

Tools of Spatial Pathology: Beyond Imaging—Toward a Deeper Understanding of Tissue

Spatial pathology is not just a technological upgrade—it’s a new language for understanding disease. And this language is written with tools that don't just visualize, but reconstruct, segment, and correlate data with clinical meaning. Among them, Visium and CosMx have emerged as foundational pillars of this transformation.

Visium: Spatial Transcriptomics with Increasing Precision

Visium, from 10x Genomics, introduced a disruptive model: transforming a histological section into a molecular coordinate map. By capturing mRNA onto barcoded spots across the slide, it generates a spatial gene expression map. In its standard version, each spot covers dozens of cells. But with Visium HD, the scale changes: from 55 µm (~5,000 features) to over 11 million features at 2 µm resolution across a 6.5x6.5 mm capture area. This enables near single-cell resolution, while remaining compatible with FFPE, fresh frozen, and fixed frozen samples (1).

Most importantly, it’s already being applied in real-world pathology. In colorectal cancer, Visium HD has allowed precise mapping of tumor borders, identifying distinct immune cell populations—macrophages, T cells—and their differential expression at the tumor edge, reshaping how we interpret immune infiltration and the tumor microenvironment (5).

CosMx: In Situ Imaging with Subcellular Resolution

CosMx, from NanoString, offers a complementary approach: highly multiplexed in situ imaging. With the ability to simultaneously visualize over 6,000 RNA and 64 proteins in a single tissue section—while preserving morphology and achieving subcellular resolution—this platform stands out (6).

It scans the tissue with surgical precision, segments it cell by cell, and produces ultra-high-resolution cellular atlases. This means we can now observe the coexistence of multiple cellular identities in situ, quantify dozens of biomarkers within their native spatial context, and analyze both frozen and FFPE samples. It’s not just a leap in imaging—it’s a leap in knowledge.

3D Models: From Slide Readers to Volume Explorers

But the next level is volumetric. 3D pathology breaks away from the slide-based paradigm. Using techniques like light-sheet microscopy or microCT, we can scan intact tissue and reconstruct it as a continuous volume. This eliminates sub-sampling bias and reveals the real continuity of key structures—vascular networks, invasion patterns, and compromised margins.

The hypothesis is clear: more volume, better prediction. Platforms like TriPath are already proving it. Combining 3D microscopy with deep learning, TriPath analyzes entire tissue volumes and predicts clinical outcomes with greater accuracy than traditional 2D methods. By incorporating the actual tumor volume, it reduces sampling bias and enhances prognostic power (7).

AI and Cellular Neighborhoods

And while Visium and CosMx generate the data, tools like CytoCommunity help interpret it (8). These algorithms identify condition-specific tissue cellular neighborhoods (TCNs) by analyzing the spatial organization of cellular phenotypes. It has already shown clinical relevance in colorectal and breast cancer, uncovering spatial patterns tied to risk stratification and treatment response.

We are entering a new dimension of pathology—one where every cell has a location, a coordinate, and a role within its ecosystem. Visium, CosMx, and 3D platforms are not just tools: they are the gateway to a spatially-aware digital pathology. And that shift is already underway.

Now… with all this technology in our hands, it’s time to take on the next challenge.

Learning to think in 3D: the challenge of the tumor microenvironment.

Perhaps the most promising aspect of spatial pathology—and I have to be honest, it’s what excites me the most—is that we are now being challenged to think in 3D. Not just to interpret histologic patterns, but to diagnose based on the microenvironment.

As pathologists, we all know that what we see is just a “snapshot” of what happened in the tissue. But like any living system, the microenvironment is dynamic and structured. It's no longer just about which cells are present, but where they are, who they’re interacting with, and what signals they’re exchanging.

Spatial tools allow us to detect patterns that traditional histology simply cannot see. One striking example: a newly described functional T cell state (TSTR) has been mapped in several cancers , increasing after immunotherapy—especially in patients who don’t respond well (9). Its distribution within the tumor—its “immune neighborhood”—may help explain why certain treatments fail. A conventional analysis would have missed that.

The tumor microenvironment is a landscape of tension and balance. Immune cells, fibroblasts, and vessels coexist in configurations that can either support or suppress tumor growth. High-resolution spatial technologies have already revealed specific “cellular neighborhoods”: for example, tumor-associated macrophages (TAMs) clustering around malignant glands—linked to poor prognosis—or distinct neutrophil subtypes distributed in patterns associated with tumor progression (10).

And it doesn’t stop there. The integration of spatial transcriptomics with 3D extracellular matrix (ECM) imagingenables us to map cellular states within complex three-dimensional niches (11). Niches like the epithelial-mesenchymal transition (EMT) zone or the dendritic cell niche can now be explored through 3D ligand-receptor signaling, revealing mechanisms of tumor progression, ECM remodeling, and immune escape. Understanding these spatial trajectories—like how a tumor transitions from “cold” to “hot,” or how tertiary lymphoid structures (TLS) emerge—is critical for anticipating clinical evolution.

Amid this sea of data, artificial intelligence becomes an indispensable ally.

The amount of information generated by spatial pathology is overwhelming: tens of thousands of cells, hundreds of variables, deeply layered maps. No human eye can process that alone. That’s where deep learning comes in. Models have already been developed that analyze spatial architecture in lung cancer and outperform traditional biomarkers in predicting treatment response. Other algorithms identify spatial histologic patterns in kidney tissue, delivering highly accurate prognostics for nephropathies.

A recent breakthrough is SpaLinker, a framework that integrates spatial transcriptomics with massive omics data to link tumor microenvironment (TME) features to clinical phenotypes. SpaLinker can map specific TME architectures—cell distributions, immune infiltration, TLS formation, tumor-associated neutrophils—and connect them to outcomes across multiple cancer types. This opens a new door: not just to understand tumors, but to predict their behavior based on their spatial architecture (12).

AI doesn’t just accelerate. It integrates. It connects digital images, molecular data, and clinical context into a holistic view of the tumor. And in the not-so-distant future, it will compare sequential biopsies in real time and suggest the next therapeutic step. In this new three-dimensional world, AI becomes the pathologist’s compass.

But no matter how promising or exciting it sounds… there's always a catch—
and this discipline is no exception.

Challenges for Clinical Adoption: Costs and Regulation

Cost and economic models

The implementation of spatial omics technologies currently demands a significant capital investment. The cost of platforms and associated workflows ranges between $300,000 and $1 million (13), not including the added expenses of specialized reagents and intensive computational analysis. As a result, the per-sample cost can reach several thousand dollars—prohibitive for many hospitals, especially those with limited budgets or without grant support.

Beyond the initial investment, there is an inherent tension between technical performance and economic feasibility. High-plex, high-resolution platforms—capable of analyzing hundreds of genes or proteins simultaneously—are intrinsically more expensive, whereas more affordable solutions often require compromising on depth or precision. This presents a critical dilemma for clinical adoption: how to ensure sufficient performance without jeopardizing economic sustainability.

In this context, early clinical applications are likely to focus on high-value medical scenarios, where even limited multiplexing may provide significant diagnostic, prognostic, or therapeutic insights (e.g., rare diseases, difficult tumors, or costly treatment decisions, a standard set of prognostically relevant transcriptomic signatures with real clinical impact is needed).

Still, there are encouraging signs: technological maturation, automation, and growing market competition are driving a progressive reduction in costs (13). Companies like 10x Genomics are already offering more integrated solutions—such as Visium slides compatible with high-resolution microscopes—aimed at lowering the barrier to entry (13). Similarly, innovations like Well-ST-seq, which promise cell-level spatial transcriptomics with reduced RNA and reagent consumption, may significantly cut down expenses in the near future (14).

The integration of artificial intelligence, machine learning, and cloud-based analytics also contributes to improving scalability and processing efficiency, enabling more data to be analyzed per dollar and reducing computational bottlenecks (15).

From an economic standpoint, the key argument should not be limited to lowering the unit cost of the test, but rather to demonstrating clinical and systemic value. For instance, a molecular test that costs $200 but prevents an ineffective $20,000 treatment not only justifies its use—it reframes the logic of healthcare spending. It is critical to shift from a focus on direct test costs to a comprehensive evaluation of the patient journey, considering quality of life, avoidance of unnecessary toxicity, and overall system efficiency (16).

In the short term, strategies like centralizing processing in specialized reference laboratories, using targeted panels(rather than whole-transcriptome approaches), or gradual adoption in centers of excellence can improve economic feasibility. Over the longer term, the synergy between industry and academia, the expanding biopharma market, and the development of alternative business models (e.g., service-based specialty labs) could represent additional ways to facilitate access.

Validation, Regulation, and Reimbursement: Turning Innovation into Clinical Reality

No diagnostic technology enters clinical practice by magic. It needs rigorous validation, regulatory support, and just as crucial reimbursement pathways (which can vary widely across countries). Spatial pathology will be no exception.

Currently used mainly for research, these technologies must prove — with robust data — that they’re reliable, reproducible, and clinically useful. That means head-to-head comparisons with current standards, clearly defined performance metrics, and consistent results across clinical settings. In high-stakes decisions, methodological rigor isn’t optional — it’s essential.

Regulatory bodies like the FDA and EMA are moving toward specific guidelines, but at different speeds. While the FDA shows flexibility (e.g., accepting real-world data), the EMA often demands more clinical evidence and longer follow-up (17). This mismatch creates a regulatory bottleneck that slows adoption, increases costs, and fuels uncertainty. Global harmonization isn’t a dream — it’s a strategic need.

Standardization is another critical challenge. Each platform (Visium, CosMx, etc.) follows different protocols, making validation and reproducibility difficult. We need universal quality controls, automated workflows, and reference materials. And it goes beyond the wet lab: standardizing data pipelines, analysis methods, and reporting is just as urgent. FAIR principles and open science will be key to building clinical trust (18).

Spatial omics will likely enter practice first as complementary tests (LDTs) in specialized labs — a transitional bridge that allows evidence-building without full regulatory approval. But no matter how good the test is, it won’t reach patients without reimbursement. Today, spatial tests often lack CPT codes or insurer coverage (16). Moving toward value-based reimbursement — where avoided costs and improved outcomes are recognized — is essential. PLA codes from the AMA are a good start, but more systemic solutions are needed(19).

Finally, we must address data governance. Spatial data, especially when genomic, can be sensitive and identifiable. Ensuring privacy without blocking innovation will require strong frameworks (18). Federated learning and blockchain may help — but clear rules and consensus will be key.

Toward a Subspecialty of “Spatial Pathology”?

As spatial pathology begins to establish itself within clinical and academic practice, a timely question is emerging:
Are we witnessing the birth of a new subspecialty?
Should we start formally training — and recognizing — a new kind of expert: the spatial pathologist?

It may sound disruptive, but it wouldn’t be the first time. In recent decades, we’ve seen the rise of molecular pathology, informatics, and digital pathology — all born from the need to adapt to new technologies and data streams. Today, the explosion of spatial platforms capable of generating high-resolution molecular maps of tissues presents us with an even deeper transformation: not just new tools, but a new way of understanding disease.

We are not merely innovating — we are shifting the diagnostic paradigm.

Spatial pathology doesn't just change what we look at, but how we interpret it. It moves us beyond pattern recognition in two dimensions, pushing us to reconstruct tissue interactions, identify functional gradients, and map immune or clonal landscapes in 3D. To do this, we need new skills that go far beyond morphology — we need knowledge in genomics, transcriptomics, computational biology, and data science.

This is where the figure of the spatial pathologist comes in:
A hybrid professional with the trained eye of a classical pathologist, expanded by the ability to integrate multiple layers of biological information into a cohesive clinical story. They are not simply observers of the tissue — they are readers of context, architects of the tumor microenvironment, synthesizing morphology, multi-omics, and spatial data into actionable insights.

Let’s consider a real scenario:
In a major cancer center, a multidisciplinary team discusses a challenging case. The spatial pathologist doesn’t just report morphology — they provide a functional tissue map, pinpointing immune niches, tumor escape routes, or metabolic vulnerabilities. Their input can reframe therapeutic strategies, reduce ineffective treatments, or even predict resistance patterns.This isn’t science fiction. It’s already happening in pioneering institutions.

But for this future to scale, we need to transform how we educate and train pathologists. For decades, pathology has been taught through a 2D lens — glass slides and morphological patterns. This has produced strong diagnosticians, but often without the training to handle large-scale data or computational tools. It’s not their fault — it’s a systemic gap.

Curricular reform is no longer optional — it’s urgent.
Residency programs must integrate digital pathology, spatial omics, image analysis, data workflows, and foundational programming (R, Python) from day one. Bioinformatics should not be an optional skill for “tech-savvy” individuals, but a core competency.

Of course, this is not just a technical shift — it's a cultural one. Transitioning from classical pathology to a data-integrated model means challenging mental models, habits, and sometimes fear of the unknown. But if our history has taught us anything, it’s that pathology is resilient and adaptable. We embraced immunohistochemistry, digital scanning, molecular profiling… and now, we’re ready for the next leap.

Does this mean every pathologist must become a spatial specialist?
Not necessarily. But it does mean that all pathologists — regardless of their subfield — will need to understand the principles of spatial data, how to interpret it, and how to use it to guide clinical decisions. Integration must become part of our shared professional language.

The future of pathology will be — by necessity — multidisciplinary.
Translating spatial data into clear, clinically relevant reports will require tight collaboration between pathologists, data scientists, bioinformaticians, and clinicians. And here, the pathologist will not be replaced, but repositioned as the key integrator — the one who understands not just the cell, but the system it lives in.

Ultimately, spatial pathology is shifting our gaze.
We are moving from isolated cells to dynamic ecosystems, from static slides to functional maps, from morphology alone to multimodal integration. If we can overcome today’s challenges — cost, validation, training, and cultural resistance — the spatial pathologist may soon become not just a concept, but a recognized, essential figure in precision medicine.

And then, that phrase we love to say — “the pathologist is an architect of tissue” —
will no longer be a metaphor.
It will be an accurate, everyday description of who we are and what we do.

---

References

1. Maciejewski K, Czerwinska P. Scoping Review: Methods and Applications of Spatial Transcriptomics in Tumor Research. Cancers. 2024 Jan;16(17):3100.

2. Laleo, Por el Amor a la Medicina. [Internet]. 2025 [cited 2025 Jul 2]. El futuro de la histología. Available from: https://www.laleo.com/histologia-texto-y-atlas-correlacion-con-la-biologia-molecular-y-celular-p-204496.html

3. Liu JT, Chow SS, Colling R, Downes MR, Farré X, Humphrey P, et al. Engineering the future of 3D pathology. J Pathol Clin Res. 2023 Nov 2;10(1):e347.

4. Stoltzfus CR, Reder N, Alvarsson A, DeWitt A, Storey C, Wilson J, et al. Quantitative 3D assessment of the lung cancer microenvironment using multi-resolution open-top light-sheet microscopy. JCO. 2023 Jun 1;41(16_suppl):2616–2616.

5. Oliveira MFD, Romero JP, Chung M, Williams SR, Gottscho AD, Gupta A, et al. High-definition spatial transcriptomic profiling of immune cell populations in colorectal cancer. Nat Genet. 2025 Jun;57(6):1512–23.

6. CosMx SMI - Single-Cell Imaging [Internet]. NanoString. 2025 [cited 2025 Jul 2]. Available from: https://nanostring.com/products/cosmx-spatial-molecular-imager/single-cell-imaging-overview/

7. Song AH, Williams M, Williamson DFK, Chow SSL, Jaume G, Gao G, et al. Analysis of 3D pathology samples using weakly supervised AI. Cell. 2024 May 9;187(10):2502-2520.e17.

8. Hu Y, Rong J, Xu Y, Xie R, Peng J, Gao L, et al. Unsupervised and supervised discovery of tissue cellular neighborhoods from cell phenotypes. Nat Methods. 2024 Feb;21(2):267–78.

9. Chu Y, Dai E, Li Y, Han G, Pei G, Ingram DR, et al. Pan-cancer T cell atlas links a cellular stress response state to immunotherapy resistance. Nat Med. 2023 Jun;29(6):1550–62.

10. Lee RY, Ng CW, Rajapakse MP, Ang N, Yeong JPS, Lau MC. The promise and challenge of spatial omics in dissecting tumour microenvironment and the role of AI. Front Oncol [Internet]. 2023 May 1 [cited 2025 Jul 2];13. Available from: https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1172314/full

11. Pentimalli TM, Schallenberg S, León-Periñán D, Legnini I, Theurillat I, Thomas G, et al. Combining spatial transcriptomics and ECM imaging in 3D for mapping cellular interactions in the tumor microenvironment. Cell Systems. 2025 May;16(5):101261.

12. Cheng X, Tang C, Dong K, You Y, Zhao X, Duan B, et al. SpaLinker identifies phenotype-associated spatial tumor microenvironment features by integrating bulk and spatial sequencing data. Cell Genomics. 2025 Jun;100893.

13. Spatial omics – microenvironments in full resolution [Internet]. [cited 2025 Jul 4]. Available from: https://kpmg.com/us/en/articles/2022/spatial-omics.html

14. Yu N, Jin Z, Liang C, Zhang J, Yang B. Well-ST-seq: cost-effective spatial transcriptomics at cellular level and high RNA capture efficiency [Internet]. bioRxiv; 2023 [cited 2025 Jul 4]. p. 2023.06.28.546974. Available from: https://www.biorxiv.org/content/10.1101/2023.06.28.546974v1

15. Spatial OMICS Market Size & Share | Industry Report, 2030 [Internet]. [cited 2025 Jul 4]. Available from: https://www.grandviewresearch.com/industry-analysis/spatial-omics-market

16. Quinn B, Llp FH. Current Issues and Options: Coverage and Reimbursement for Complex Molecular Diagnostics.

17. FDA vs. EMA: Navigating Divergent Regulatory Expectations for Cell and Gene Therapies. What Biopharma Companies Need to Know | Cromos Pharma [Internet]. 2025 [cited 2025 Jul 4]. Available from: https://cromospharma.com/fda-vs-ema-navigating-divergent-regulatory-expectations-for-cell-and-gene-therapies-what-biopharma-companies-need-to-know/

18. Pentimalli T, Karaiskos N, Rajewsky N. Challenges and Opportunities in the Clinical Translation of High-Resolution Spatial Transcriptomics. Annual review of pathology. 2024 Oct 30;20.

19. Modulim Granted CPT Code from the American Medical Association for its Noninvasive Microvascular Assessment Technology – Modulim [Internet]. [cited 2025 Jul 4]. Available from: https://modulim.com/modulim-granted-cpt-code-from-the-american-medical-association-for-its-noninvasive-microvascular-assessment-technology