Context.—:Advances in computer vision have fueled the development of artificial intelligence (AI)-based algorithms for pathology. AI-assisted approaches may streamline the diagnostic workflow and reduce variability. Objective.—:To assess the impact of an AI-assist model for human epidermal growth factor receptor 2 (HER2) scoring on pathologist reproducibility and accuracy and to understand pathologist-model interactions. Design.—:An AI-Assist algorithm for HER2 scoring, AI-Measurement of HER2 (AIM-HER2), was developed to generate slide-level scores of HER2 immunohistochemistry (IHC) aligned with guidelines from the American Society of Clinical Oncology/College of American Pathologists. AIM-HER2 was assessed in a retrospective reader study wherein HER2-trained pathologists (n = 20) scored breast cancer cases (n = 200) with and without model assistance using a 2-cohort crossover design with a 3-week washout. A separate panel of expert pathologists (n = 5) provided manual reference scores. Results.—:As an AI-assist tool, AIM-HER2 improved interrater agreement both overall and at the 0/1+ and 1+/2+ cutoffs and significantly increased positive percentage agreement at the 0/1+ and 1+/2+ cutoffs. Pathologists displayed a wide range of model override rates, and the quality of these overrides was correlated with each pathologist's manual accuracy. Measurements of AIM-HER2 accuracy were highly dependent on reference panel composition. Conclusions.—:The use of AI-assist tools, such as AIM-HER2, for scoring HER2 IHC in breast cancer may improve pathologist reproducibility and accuracy, particularly at the 0/1+ and 1+/2+ cutoffs. However, improved consistency of pathologist interpretation of AI-assisted IHC scoring guidance may be necessary for AI-assist tools to reach their full potential.

Context.-Generative artificial intelligence (AI) has emerged as a transformative force in various fields, including anatomic pathology, where it offers the potential to significantly enhance diagnostic accuracy, workflow efficiency, and research capabilities. Objective.-To explore the applications, benefits, and challenges of generative AI in anatomic pathology, with a focus on its impact on diagnostic processes, workflow efficiency, education, and research. Data Sources.-A comprehensive review of current literature and recent advancements in the application of generative AI within anatomic pathology, categorized into unimodal and multimodal applications, and evaluated for clinical utility, ethical considerations, and future potential. Conclusions.-Generative AI demonstrates significant promise in various domains of anatomic pathology, including diagnostic accuracy enhanced through AI-driven image analysis, virtual staining, and synthetic data generation; workflow efficiency, with potential for improvement by automating routine tasks, quality control, and reflex testing; education and research, facilitated by AI-generated educational content, synthetic histology images, and advanced data analysis methods; and clinical integration, with preliminary surveys indicating cautious optimism for nondiagnostic AI tasks and growing engagement in academic settings. Ethical and practical challenges require rigorous validation, prompt engineering, federated learning, and synthetic data generation to help ensure trustworthy, reliable, and unbiased AI applications. Generative AI can potentially revolutionize anatomic pathology, enhancing diagnostic accuracy, improving workflow efficiency, and advancing education and research. Successful integration into clinical practice will require continued interdisciplinary collaboration, careful validation, and adherence to ethical standards to ensure the benefits of AI are realized while maintaining the highest standards of patient care.

Abstract Understanding the role of the tumor microenvironment (TME) in lung cancer is critical to improving patient outcomes. We identified four histology-independent archetype TMEs in treatment-naïve early-stage lung cancer using imaging mass cytometry in the TRACERx study (n = 81 patients/198 samples/2.3 million cells). In immune-hot adenocarcinomas, spatial niches of T cells and macrophages increased with clonal neoantigen burden, whereas such an increase was observed for niches of plasma and B cells in immune-excluded squamous cell carcinomas (LUSC). Immune-low TMEs were associated with fibroblast barriers to immune infiltration. The fourth archetype, characterized by sparse lymphocytes and high tumor-associated neutrophil (TAN) infiltration, had tumor cells spatially separated from vasculature and exhibited low spatial intratumor heterogeneity. TAN-high LUSC had frequent PIK3CA mutations. TAN-high tumors harbored recently expanded and metastasis-seeding subclones and had a shorter disease-free survival independent of stage. These findings delineate genomic, immune, and physical barriers to immune surveillance and implicate neutrophil-rich TMEs in metastasis. Significance: This study provides novel insights into the spatial organization of the lung cancer TME in the context of tumor immunogenicity, tumor heterogeneity, and cancer evolution. Pairing the tumor evolutionary history with the spatially resolved TME suggests mechanistic hypotheses for tumor progression and metastasis with implications for patient outcome and treatment. This article is featured in Selected Articles from This Issue, p. 897

Supplementary Figures 1-22 with the corresponding figure legends inline. Supplementary Figure S1. TRACERx 100 imaging mass cytometry cohort. Supplementary Figure S2. Analysis of imaging mass cytometry data. Supplementary Figure S3. Characterisation of cell subtypes and spatial cellular communities in non-small cell lung cancer. Supplementary Figure S4. Clinicogenomic correlations with cell subtypes and communities. Supplementary Figure S5. TME class associations with cell types and clinical variables. Supplementary Figure S6. Cancer cell-intrinsic and -extrinsic features associated with immune cell infiltration. Supplementary Figure S7. Spatial, histological and metabolic features of the Tumour/Stroma:Neutrophil High TME class. Supplementary Figure S8. Transcriptomic features of TS:Neutrophil High TME class and Tumour-Associated Neutrophil scoring. Supplementary Figure S9. Somatic mutations in PIK3CA were associated with neutrophil recruitment through CXCL8 upregulation. Supplementary Figure S10. TAN score association with disease-free survival. Supplementary Figure S11. Validation of the prognostic association from the TAN scoring approach with an automated, deep learning approach in The Cancer Genome Atlas. Supplementary Figure S12. Imaging mass cytometry panel development. Supplementary Figure S13. Spillover matrices for imaging mass cytometry data. Supplementary Figure S14. Multiplexed Consensus Cell Segmentation. Supplementary Figure S15. Investigation of batch effects. Supplementary Figure S16. Raw pixel intensities. Supplementary Figure S17. Pathologist-guided labels. Supplementary Figure S18. Communities methodology and histology associations. Supplementary Figure S19. Multiplexed immunohistochemistry validation of checkpoint molecule expression. Supplementary Figure S20. Cell-cell relationships differ by TME class. Supplementary Figure S21. Cell-cell relationships differ by tumour genomics. Supplementary Figure S22. Cell-cell relationships based on tumour cell phenotypes.

IMC workflow defines the single-cell spatial landscape of the NSCLC tumor microenvironment. A, TRACERx 100 IMC cohort. We developed and applied two IMC antibody panels, Pan-immune and T cells and stroma, to tissue microarrays (TMA) from clinical samples collected at surgical resection (created with BioRender.com). B, Targets of antibodies described in this study. Bold text indicates targets detected in both IMC panels. C, IMC data were acquired from stained TMAs and processed to identify single cells and their phenotypes. D, 40,000 μm2 crops of IMC images representing the markers from B with corresponding cell types from the pan-immune panel, unless annotated with an asterisk for the T cells and stroma panel only. E, A heat map of the z-score normalized median intensities of markers from the pan-immune panel across the identified cell subtypes. F, Proportion of major immune cell types identified in the pan-immune IMC data set per TMA core, calculated over the total tissue area (illustrated as blue and gold domains), tumor/epithelial compartment (gold domain), or the stromal compartment (blue domain). In two normal cores, the epithelial cell signal reflected very thin cells, which were not resolved into an epithelial compartment. All data from these cores are represented by the stroma compartment. Cell types color legend applies to D and F, where asterisks denote cell types identified in T cells and stroma panel only. LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; NSCLC, non–small cell lung cancer; other, other non–small cell lung cancer histologies; IMC, imaging mass cytometry.

Pathology is the study of microscopic inspection of tissue, and a pathology diagnosis is often the medical gold standard to diagnose disease. Pathology images provide a unique challenge for computer-vision-based analysis: a single pathology Whole Slide Image (WSI) is gigapixel-sized and often contains hundreds of thousands to millions of objects of interest across multiple resolutions. In this work, we propose PathoLogy Universal TransfOrmer (PLUTO): a light-weight pathology FM that is pre-trained on a diverse dataset of 195 million image tiles collected from multiple sites and extracts meaningful representations across multiple WSI scales that enable a large variety of downstream pathology tasks. In particular, we design task-specific adaptation heads that utilize PLUTO's output embeddings for tasks which span pathology scales ranging from subcellular to slide-scale, including instance segmentation, tile classification, and slide-level prediction. We compare PLUTO's performance to other state-of-the-art methods on a diverse set of external and internal benchmarks covering multiple biologically relevant tasks, tissue types, resolutions, stains, and scanners. We find that PLUTO matches or outperforms existing task-specific baselines and pathology-specific foundation models, some of which use orders-of-magnitude larger datasets and model sizes when compared to PLUTO. Our findings present a path towards a universal embedding to power pathology image analysis, and motivate further exploration around pathology foundation models in terms of data diversity, architectural improvements, sample efficiency, and practical deployability in real-world applications.

Spatial features associated with neoantigen burden and immune low TMEs. A, Correlation of densities of spatial cellular communities and the burden of expressed clonal, subclonal and total neoantigens predicted to bind intact HLA alleles, after accounting for HLA LOH, in LUAD (n = 31, 51 tumor cores) and LUSC (n = 17, 37 tumor cores). Bar plot shows the median neoantigen burden with whiskers extending to the 75th percentile. B, Comparison of the densities of spatial cellular communities in a given TME class compared with all other TME classes combined. LUAD: n = 21 TS:TIL+MΦ high cores, n = 20 T:TIL+MΦ excluded cores, n = 13 TS:Immune low cores, n = 11 TS:Neutrophil high cores. LUSC: n = 13 TS:TIL+MΦ high cores, n = 12 T:TIL+MΦ excluded cores, n = 8 TS:Immune low cores, n = 15 TS:Neutrophil high cores. Box sizes in A and B correspond to T values. C, Community and cell subtype maps from a LUAD tumor core with a high burden of expressed clonal neoantigens and high densities of C2:T-cell enriched and C6:macrophage and T cells communities. D, Community and cell subtype maps from a LUSC tumor core with a high burden of expressed clonal neoantigens and high densities of community C9:B cells and plasma cells. Single cells in C and D are colored by community according to the color legend below D or cell subtype as indicated. Scale bars, 200 μm. Middle, an enlargement of the area highlighted with a white box in the left plot with matched cell subtypes shown in the right plot. E, Schematic of αSMA+ fibroblast barrier score calculation. The barrier score measures the degree of spatial interpositioning of tumor cell–adjacent αSMA+ fibroblasts between CD8 T cells and their nearest tumor cell(s) in a tissue core. In the lower half of the schematic, three nearest tumor cells are defined for the green CD8 T cell, all six hops away. Tumor cell–adjacent αSMA+ fibroblasts are found on two of these three paths from CD8 T-cell to tumor cell, resulting in a barrier score of ⅔. F, Boxplot comparing the αSMA+ fibroblast barrier scores in a given TME class compared with all other TME classes combined in LUAD (n = 36, 57 tumor cores) and LUSC (n = 22, 45 tumor cores). Boxplots show median and lower and upper quartile values, and whiskers extend up to 1.5 × IQR above and below the quartiles. G, Representative IMC images and cell type maps from LUAD and LUSC tumor cores classified as TS:Immune low with a high barrier score. Scale bars, 200 μm. P values in A, B, and F and T values in A and B were calculated in a linear mixed-effects model with patient as a random effect, using smoking status as a fixed effect in A with a P value < 0.05 considered significant. LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; panCK, pancytokeratin; TS, tumor/stroma; T, tumor; TIL, tumor-infiltrating lymphocytes; MΦ, macrophage; *, P < 0.05; **, P < 0.01.

All consortium members and their affiliations.

Supplementary Table S1. Imaging mass cytometry antibody panel information. Supplementary Table S2. TRACERx 100 imaging mass cytometry cohort information. Supplementary Table S3. TRACERx 421 cohort information. Supplementary Table S4. Median cell densities per histology, tissue type and TME class. Supplementary Table S5. Multivariable model of disease-free survival including TAN score, relevant clinical features together with PIK3CA mutation status (a) or recent subclonal expansion (b).

Neutrophil-rich TMEs are associated with activating mutations in PI3K and tumor-intrinsic CXCL8 upregulation. A, Representative crops of tumor-level H&E images with low TAN scores in the tumor nest and stroma (left) and high TAN scores in the tumor nest and stroma (right), inferred as the proportion of the neutrophil area in tumor/stroma from the total tumor/stroma tissue area. Scale bar, 50 μm; 400× magnification. B, Neutrophil cell density as defined by IMC compared between region-level TAN-low versus TAN-high tumor cores based on H&E scores in LUAD and LUSC. C and D, Proportion of tumor cores with (mut) and without (wt) PIK3CA driver mutations compared between TS:Neutrophil high versus other TME classes combined (C) and region-level TAN-High versus TAN-Low cores (D) in LUSC. P values were derived from a Chi-square test. E, Neutrophil cell density by PIK3CA mutation status, points colored by TME class assignment. F and G, TMM expression values for CXCL8 compared by PIK3CA mutation status (F) and between TME classes (G) in LUSC. H, Immunofluorescence images of CXCL8 RNAscope multiplexed with antibody staining of pancytokeratin (panCK) or MPO in an LUSC tumor region with a TS:Neutrophil high TME and subclonal PIK3CA mutation, and an LUSC patient with multiple TAN-high tumor regions and a clonal PIK3CA mutation. panCK and MPO examples for CRUK0075:R2 illustrate the same region of interest, whereas different regions of interest are shown for CRUK0468:R6. Scale bar, 100 μm. P values for B and E were calculated in a linear mixed-effects model with patient as the random-effect covariate. P values in F and G were derived from a limma–voom differential expression analysis correcting for multiple regions per tumor. ·, P < 0.1; *, P < 0.05; **, P < 0.01; ***, P < 0.001; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; TMM, trimmed mean of M-values; TS, tumor/stroma; TIL, tumor-infiltrating lymphocyte; MΦ, macrophage; TAN, tumor-associated neutrophils; H&E, hematoxylin and eosin; mt, mutant; wt, wild-type.

CONTEXT.—:Programmed death ligand-1 (PD-L1) immunohistochemistry companion diagnostic assays play a crucial role as predictive markers in patients being considered for immune checkpoint inhibitor therapy. However, because of a convergence of several factors, including recognition of increased types of cancers susceptible to immunotherapy, increasing numbers of immune checkpoint inhibitors, and release of multiple PD-L1 immunohistochemistry antibodies with differing reporting systems, this complex testing environment has led to significant levels of confusion for pathologists and medical oncologists.OBJECTIVE.—:To identify which processes and procedures have contributed to the current challenges surrounding programmed death receptor-1 (PD-1)/PD-L1 companion diagnostics and to propose potential remedies to this issue. This is based upon input from key industrial stakeholders in conjunction with the College of American Pathologists Personalized Health Care Committee.DESIGN.—:A meeting of representatives of pharmaceutical and in vitro diagnostic companies along with the Personalized Health Care Committee reviewed the process of release of the PD-L1 companion diagnostic assays using a modified root cause analysis format. The modified root cause analysis envisioned an ideal circumstance of development and implementation of a companion diagnostic to identify shortcomings in the rollout of the PD-L1 assay and to suggest actions to improve future companion diagnostic assay releases.RESULTS.—:The group recommended improvements to key principles in companion diagnostics implementation related to multi-stakeholder communication, increased regulatory flexibility to incorporate postapproval medical knowledge, improved cross-disciplinary information exchange between medical oncology and pathology societies, and enhanced postmarket training programs.CONCLUSIONS.—:The rapidly changing nature of and increasing complexity associated with companion diagnostics require a fundamental review of processes related to their design, implementation, and oversight.

PDF file - 59K, Pharmacodynamic Parameters

PDF file - 72K, Most Frequent Adverse Events (≥ 15%, any dose group) by Initial Dose Group (Safety Population)

CONTEXT.—:Neurotrophic receptor tyrosine kinase (NTRK) fusion testing has both diagnostic and therapeutic implications for patient care. With 2 tumor-agnostic US Food and Drug Administration-approved tropomyosin receptor kinase (TRK) inhibitors, testing is increasingly used for therapeutic decision making. However, the testing landscape for NTRK fusions is complex, and optimal testing depends on the clinicopathologic scenario.OBJECTIVE.—:To compare different NTRK testing methods to help pathologists understand test features and performance characteristics and make appropriate selections for NTRK fusion detection for their laboratory and individual patient specimens.DATA SOURCES.—:A literature search for NTRK gene fusions and TRK protein was performed, including papers that discussed treatment, testing methodology, and detection or prevalence of fusion-positive cases.CONCLUSIONS.—:As standard of care in some tumor types, next-generation sequencing (NGS) panel testing is a cost effective and reliable way to detect a broad range of NTRK fusions. The design of the panel and use of DNA or RNA will affect performance characteristics. Pan-TRK immunohistochemistry may be used as a rapid, less expensive screen in cases that will not undergo routine NGS testing, or on specimens unsuitable for NGS testing. Fluorescence in situ hybridization may be appropriate for low-tumor-content specimens that are unsuitable for NGS testing. Quantitative reverse transcription polymerase chain reaction is best suited for monitoring low-level disease of a specific, previously identified target. This information should help laboratories develop a laboratory-specific NTRK testing algorithm that best suits their practice setting and patients' needs.

PDF file - 56K, Mean Pharmacokinetic Parameters

PDF file - 50K, Primary Antibodies Used for Immunohistochemical Staining

PDF file - 193K, Pre- and on-therapy immunohistochemistry of paired biomarkers in skin (p-EGFR, p-MAPK, Ki67), vascular (p-MAPK, Ki67), and tumor (p-EGFR, p-MAPK, p-Akt and Ki67) samples in two patients treated at 25 and 550 mg, respectively