Application of digital pathology‐based advanced analytics of tumour microenvironment organisation to predict prognosis and therapeutic response DOI Creative Commons
Xiao Fu, Erik Sahai, Anna Wilkins

et al.

The Journal of Pathology, Journal Year: 2023, Volume and Issue: 260(5), P. 578 - 591

Published: Aug. 1, 2023

Abstract In recent years, the application of advanced analytics, especially artificial intelligence (AI), to digital H&E images, and other histological image types, has begun radically change how images are used in clinic. Alongside recognition that tumour microenvironment (TME) a profound impact on phenotype, technical development highly multiplexed immunofluorescence platforms enhanced biological complexity can be captured TME with high precision. AI an increasingly powerful role quantitation features association such clinically important outcomes, as occurs distinct stages conventional machine learning. Deep‐learning algorithms able elucidate patterns inherent input data minimum levels human and, hence, have potential achieve relevant predictions discovery features. Furthermore, diverse repertoire deep‐learning interrogate extends beyond convolutional neural networks include attention‐based models, graph networks, multimodal models. To date, models largely been evaluated retrospectively, outside well‐established rigour prospective clinical trials, part because traditional trial methodology may not always suitable for assessment technology. However, enable pathology‐based analytics meaningfully care, specific measures ‘added benefit’ current standard care validation setting important. This will need accompanied by adequate explainability interpretability. Despite challenges, combination expanding datasets, increased computational power, possibility integration pre‐clinical experimental insights into model means there is exciting future progress these applications. © 2023 The Authors. Journal Pathology published John Wiley & Sons Ltd behalf Pathological Society Great Britain Ireland.

Language: Английский

SpiDe-Sr: blind super-resolution network for precise cell segmentation and clustering in spatial proteomics imaging DOI Creative Commons
Rui Chen,

Jiasu Xu,

Boqian Wang

et al.

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: March 28, 2024

Abstract Spatial proteomics elucidates cellular biochemical changes with unprecedented topological level. Imaging mass cytometry (IMC) is a high-dimensional single-cell resolution platform for targeted spatial proteomics. However, the precision of subsequent clinical analysis constrained by imaging noise and resolution. Here, we propose SpiDe-Sr, super-resolution network embedded denoising module IMC enhancement. SpiDe-Sr effectively resists improves 4 times. We demonstrate respectively cells, mouse human tissues, resulting 18.95%/27.27%/21.16% increase in peak signal-to-noise ratio 15.95%/31.63%/15.52% cell extraction accuracy. further apply to study tumor microenvironment 20-patient breast cancer cohort 269,556 single discover invasion Gram-negative bacteria positively correlated carcinogenesis markers negatively immunological markers. Additionally, also compatible fluorescence microscopy imaging, suggesting an alternative tool image super-resolution.

Language: Английский

Citations

7
Unsupervised Analysis of Optical Imaging Data for the Discovery of Reactivity Patterns in Metal Alloy DOI Creative Commons
Rui Li, Aleksei Makogon, Tatiana Galochkina

et al.

Small Methods, Journal Year: 2023, Volume and Issue: 7(10)

Published: June 29, 2023

Abstract Operando wide‐field optical microscopy imaging yields a wealth of information about the reactivity metal interfaces, yet data are often unstructured and challenging to process. In this study, power unsupervised machine learning (ML) algorithms is harnessed analyze chemical images obtained dynamically by reflectivity in combination with ex situ scanning electron identify cluster particles Al alloy. The ML analysis uncovers three distinct clusters from unlabeled datasets. A detailed examination representative patterns confirms communication generated OH − fluxes within particles, as supported statistical size distribution finite element modelling (FEM). procedures also reveal statistically significant under dynamic conditions, such pH acidification. results align well numerical model communication, underscoring synergy between data‐driven physics‐driven FEM approaches.

Language: Английский

Citations

14
A review on deep learning applications in highly multiplexed tissue imaging data analysis DOI Creative Commons
Mohammed Zidane, Ahmad Makky,

Matthias Bruhns

et al.

Frontiers in Bioinformatics, Journal Year: 2023, Volume and Issue: 3

Published: July 26, 2023

Since its introduction into the field of oncology, deep learning (DL) has impacted clinical discoveries and biomarker predictions. DL-driven predictions in oncology are based on a variety biological data such as genomics, proteomics, imaging data. DL-based computational frameworks can predict genetic variant effects gene expression, well protein structures amino acid sequences. Furthermore, DL algorithms capture valuable mechanistic information from several spatial “omics” technologies, transcriptomics proteomics. Here, we review impact that combination artificial intelligence (AI) with omics technologies had focusing applications biomedical image analysis, encompassing cell segmentation, phenotype identification, cancer prognostication, therapy prediction. We highlight advantages using highly multiplexed images (spatial proteomics data) compared to single-stained, conventional histopathological (“simple”) images, former provide insights cannot be obtained by latter, even aid explainable AI. reader advantages/disadvantages pipelines used preprocessing (cell type annotation). Therefore, this also guides choose pipeline best fits their In conclusion, continues established an essential tool discovering novel mechanisms when combined tissue balance medical data, role routine will become more important, supporting diagnosis prognosis enhancing decision-making, improving quality care for patients.

Language: Английский

Citations

14
Quantifying and interpreting biologically meaningful spatial signatures within tumor microenvironments DOI Creative Commons
Siyu Jing,

He-qi Wang,

Ping Lin

et al.

npj Precision Oncology, Journal Year: 2025, Volume and Issue: 9(1)

Published: March 11, 2025

The tumor microenvironment (TME) plays a crucial role in orchestrating cell behavior and cancer progression. Recent advances spatial profiling technologies have uncovered novel signatures, including univariate distribution patterns, bivariate relationships, higher-order structures. These signatures the potential to revolutionize mechanism treatment. In this review, we summarize current state of signature research, highlighting computational methods uncover spatially relevant biological significance. We discuss impact these on fundamental biology translational address challenges future research directions.

Language: Английский

Citations

0
Quantitative characterization of tissue states using multiomics and ecological spatial analysis DOI Creative Commons
Daisy Yi Ding, Zeyu Tang, Bokai Zhu

et al.

Nature Genetics, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

The spatial organization of cells in tissues underlies biological function, and recent advances profiling technologies have enhanced our ability to analyze such arrangements study processes disease progression. We propose MESA (multiomics ecological analysis), a framework drawing inspiration from concepts delineate functional shifts across tissue states. introduces metrics systematically quantify diversity identify hot spots, linking patterns phenotypic outcomes, including Furthermore, integrates single-cell multiomics data facilitate an in-depth, molecular understanding cellular neighborhoods their interactions within microenvironments. Applying diverse datasets demonstrates additional insights it brings over prior methods, newly identified structures key cell populations linked Available as Python package, offers versatile for quantitative decoding architectures omics health disease. Multiomics analysis (MESA) calculates ecodiversity-inspired spatially resolved integrated with data, enabling the comparison states range conditions.

Language: Английский

Citations

0
Exploring the single-cell immune landscape of kidney allograft inflammation using imaging mass cytometry DOI Creative Commons

Mariam P. Alexander,

Mark Zaidi, Nicholas B. Larson

et al.

American Journal of Transplantation, Journal Year: 2023, Volume and Issue: 24(4), P. 549 - 563

Published: Nov. 17, 2023

Kidney allograft inflammation, mostly attributed to rejection and infection, is an important cause of graft injury loss. Standard histopathological assessment inflammation provides limited insights into biological processes the immune landscape. Here, using imaging mass cytometry with a panel 28 validated biomarkers, we explored single-cell landscape kidney in 32 transplant biopsies 247 high-dimensional histopathology images various phenotypes (antibody-mediated rejection, T cell-mediated BK nephropathy, chronic pyelonephritis). Using novel analytical tools, for cell segmentation, segmented over 900 000 cells developed tissue-based classifier 3000 manually annotated microstructures (glomeruli, tubules, interstitium, arteries). PhenoGraph, identified 11 9 nonimmune clusters found high prevalence memory macrophage-enriched populations across phenotypes. Additionally, trained machine learning identify spatial biomarkers that could discriminate between different inflammatory Further validation larger cohorts more will likely help interrogate depth than has been possible date.

Language: Английский

Citations

11
Precise immunofluorescence canceling for highly multiplexed imaging to capture specific cell states DOI Creative Commons
Kosuke Tomimatsu,

Takeru Fujii,

Ryoma Bise

et al.

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: May 8, 2024

Abstract Cell states are regulated by the response of signaling pathways to receptor ligand-binding and intercellular interactions. High-resolution imaging has been attempted explore dynamics these processes and, recently, multiplexed profiled cell achieving a comprehensive acquisition spatial protein information from cells. However, specificity antibodies is still compromised when visualizing activated signals. Here, we develop Precise Emission Canceling Antibodies (PECAbs) that have cleavable fluorescent labeling. PECAbs enable high-specificity sequential using hundreds antibodies, allowing for reconstruction spatiotemporal pathways. Additionally, combining this approach with seq-smFISH can effectively classify cells identify their signal activation in human tissue. Overall, PECAb system serve as platform analyzing complex processes.

Language: Английский

Citations

4
Frontiers in Fluorescence imaging: Tools for the In-Situ Sensing of Disease Biomarkers DOI
Lei Yang, Hongwei Hou, Jinghong Li

et al.

Journal of Materials Chemistry B, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 3, 2024

A comprehensive overview of recent advancements in fluorescence imaging techniques for situ sensing various biomarkers, emphasizing the transformative potential artificial intelligence shaping future bioimaging.

Language: Английский

Citations

4
Spatial Patterning Analysis of Cellular Ensembles (SPACE) finds complex spatial organization at the cell and tissue levels DOI Creative Commons
Edward C. Schrom, Erin McCaffrey, Vivek Sreejithkumar

et al.

Proceedings of the National Academy of Sciences, Journal Year: 2025, Volume and Issue: 122(6)

Published: Feb. 4, 2025

Spatial patterns of cells and other biological elements drive physiologic pathologic processes within tissues. While many imaging transcriptomic methods document tissue organization, discerning these is challenging, especially when they involve multiple in complex arrangements. To address this challenge, we present Patterning Analysis Cellular Ensembles (SPACE), an R package for analysis high-plex spatial data. SPACE compatible with any data collection modality that records values (i.e., categorical cell/structure types or quantitative expression levels) at fixed coordinates 2d pixels 3d voxels). detects not only broad co-occurrence but also context-dependent associations, gradients orientations, organizational complexities. Via a robust information theoretic framework, explores all possible ensembles elements—single elements, pairs, triplets, so on—and ranks the most strongly patterned ensembles. For single images, rankings reflect differences from random assortment. sets across sample groups (e.g., genotypes, treatments, timepoints, etc.). Further tools then characterize nature each pattern intuitive interpretation. We validate demonstrate its advantages using murine lymph node images which ground truth has been defined. detect new varied datasets, including tumors tuberculosis granulomas.

Language: Английский

Citations

0
Powerful and accurate case-control analysis of spatial molecular data with deep learning-defined tissue microniches DOI Creative Commons
Yakir Reshef,

Lakshay Sood,

Michelle Curtis

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 8, 2025

Abstract As spatial molecular data grow in scope and resolution, there is a pressing need to identify key structures associated with disease. Current approaches often rely on hand-crafted features such as local abundances of manually annotated, discrete cell types, which may overlook important signals. Here we introduce variational inference-based microniche analysis (VIMA), method that combines deep learning principled statistics discover greater flexibility precision. VIMA uses autoencoder extract numerical “fingerprints” from small tissue patches capture their biological content. It these fingerprints define large number “microniches” – small, potentially overlapping groups highly similar biology span multiple samples. then rigorous microniches whose abundance correlates case-control status. We show simulations well calibrated more powerful accurate than other approaches. apply 140-gene transcriptomics dataset Alzheimer’s dementia, 54-marker CO-Detection by indEXing (CODEX) ulcerative colitis (UC), 7-marker immunohistochemistry rheumatoid arthritis (RA), each case recapitulating known identifying novel

Language: Английский

Citations

0