Augmented reality microscopy to bridge trust between AI and pathologists

Abstract Diagnostic certainty is the cornerstone of modern medicine and critical for maximal treatment benefit. When evaluating biomarker expression by immunohistochemistry (IHC), however, pathologists are hindered by complex scoring methodologies, unique positivity cut-offs and subjective staining...

Full description

Saved in:
Bibliographic Details
Main Authors: Sunil Badve, George L. Kumar, Tobias Lang, Eli Peigin, James Pratt, Robert Anders, Deyali Chatterjee, Raul S. Gonzalez, Rondell P. Graham, Alyssa M. Krasinskas, Xiuli Liu, Alexander Quaas, Romil Saxena, Namrata Setia, Laura Tang, Hanlin L. Wang, Josef Rüschoff, Hans-Ulrich Schildhaus, Khalid Daifalla, Marc Päpper, Patrick Frey, Felix Faber, Maria Karasarides
Format: Article
Language:English
Published: Nature Portfolio 2025-05-01
Series:npj Precision Oncology
Online Access:https://doi.org/10.1038/s41698-025-00899-5
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850272000801505280
author Sunil Badve
George L. Kumar
Tobias Lang
Eli Peigin
James Pratt
Robert Anders
Deyali Chatterjee
Raul S. Gonzalez
Rondell P. Graham
Alyssa M. Krasinskas
Xiuli Liu
Alexander Quaas
Romil Saxena
Namrata Setia
Laura Tang
Hanlin L. Wang
Josef Rüschoff
Hans-Ulrich Schildhaus
Khalid Daifalla
Marc Päpper
Patrick Frey
Felix Faber
Maria Karasarides
author_facet Sunil Badve
George L. Kumar
Tobias Lang
Eli Peigin
James Pratt
Robert Anders
Deyali Chatterjee
Raul S. Gonzalez
Rondell P. Graham
Alyssa M. Krasinskas
Xiuli Liu
Alexander Quaas
Romil Saxena
Namrata Setia
Laura Tang
Hanlin L. Wang
Josef Rüschoff
Hans-Ulrich Schildhaus
Khalid Daifalla
Marc Päpper
Patrick Frey
Felix Faber
Maria Karasarides
author_sort Sunil Badve
collection DOAJ
description Abstract Diagnostic certainty is the cornerstone of modern medicine and critical for maximal treatment benefit. When evaluating biomarker expression by immunohistochemistry (IHC), however, pathologists are hindered by complex scoring methodologies, unique positivity cut-offs and subjective staining interpretation. Artificial intelligence (AI) can potentially eliminate diagnostic uncertainty, especially when AI “trustworthiness” is proven by expert pathologists in the context of real-world clinical practice. Building on an IHC foundation model, we employed pathologists-in-the-loop finetuning to produce a programmed cell death ligand 1 (PD-L1) CPS AI Model. We devised a multi-head augmented reality microscope (ARM) system overlayed with the PD-L1 CPS AI Model to assess interobserver variability and gauge the pathologists’ trust in AI model outputs. Using difficult to interpret regions on gastroesophageal biopsies, we show that AI-assistance improved case agreement between any 2 pathologists by 14% (agreement on 77% vs 91%) and among 11 pathologists by 26% (agreement on 43% vs 69%). At a clinical cutoff of PD-L1 CPS ≥ 5, the number of cases diagnosed as positive by all 11 pathologists increased by 31%. Our findings underscore the benefits of fully engaging pathologists as active participants in the development and deployment of IHC AI models and frame the roadmap for trustworthy AI as a bridge to increased adoption in routine pathology practice.
format Article
id doaj-art-9c1cf13975604bb4bc5abf3a510d95ad
institution OA Journals
issn 2397-768X
language English
publishDate 2025-05-01
publisher Nature Portfolio
record_format Article
series npj Precision Oncology
spelling doaj-art-9c1cf13975604bb4bc5abf3a510d95ad2025-08-20T01:51:59ZengNature Portfolionpj Precision Oncology2397-768X2025-05-019111310.1038/s41698-025-00899-5Augmented reality microscopy to bridge trust between AI and pathologistsSunil Badve0George L. Kumar1Tobias Lang2Eli Peigin3James Pratt4Robert Anders5Deyali Chatterjee6Raul S. Gonzalez7Rondell P. Graham8Alyssa M. Krasinskas9Xiuli Liu10Alexander Quaas11Romil Saxena12Namrata Setia13Laura Tang14Hanlin L. Wang15Josef Rüschoff16Hans-Ulrich Schildhaus17Khalid Daifalla18Marc Päpper19Patrick Frey20Felix Faber21Maria Karasarides22Emory University School of MedicineBristol Myers SquibbMindpeakAugmentiqs, D.NBristol Myers SquibbJohns Hopkins University BaltimoreMD Anderson Cancer CenterEmory University School of MedicineMayo ClinicEmory University School of MedicineWashington University School of MedicineCologne University HospitalEmory University School of MedicineUniversity of ChicagoMemorial Sloan Kettering Cancer CenterUCLA David Geffen School of MedicineDiscovery Life Sciences Biomarker Services GmbHDiscovery Life Sciences Biomarker Services GmbHMindpeakMindpeakMindpeakMindpeakBristol Myers SquibbAbstract Diagnostic certainty is the cornerstone of modern medicine and critical for maximal treatment benefit. When evaluating biomarker expression by immunohistochemistry (IHC), however, pathologists are hindered by complex scoring methodologies, unique positivity cut-offs and subjective staining interpretation. Artificial intelligence (AI) can potentially eliminate diagnostic uncertainty, especially when AI “trustworthiness” is proven by expert pathologists in the context of real-world clinical practice. Building on an IHC foundation model, we employed pathologists-in-the-loop finetuning to produce a programmed cell death ligand 1 (PD-L1) CPS AI Model. We devised a multi-head augmented reality microscope (ARM) system overlayed with the PD-L1 CPS AI Model to assess interobserver variability and gauge the pathologists’ trust in AI model outputs. Using difficult to interpret regions on gastroesophageal biopsies, we show that AI-assistance improved case agreement between any 2 pathologists by 14% (agreement on 77% vs 91%) and among 11 pathologists by 26% (agreement on 43% vs 69%). At a clinical cutoff of PD-L1 CPS ≥ 5, the number of cases diagnosed as positive by all 11 pathologists increased by 31%. Our findings underscore the benefits of fully engaging pathologists as active participants in the development and deployment of IHC AI models and frame the roadmap for trustworthy AI as a bridge to increased adoption in routine pathology practice.https://doi.org/10.1038/s41698-025-00899-5
spellingShingle Sunil Badve
George L. Kumar
Tobias Lang
Eli Peigin
James Pratt
Robert Anders
Deyali Chatterjee
Raul S. Gonzalez
Rondell P. Graham
Alyssa M. Krasinskas
Xiuli Liu
Alexander Quaas
Romil Saxena
Namrata Setia
Laura Tang
Hanlin L. Wang
Josef Rüschoff
Hans-Ulrich Schildhaus
Khalid Daifalla
Marc Päpper
Patrick Frey
Felix Faber
Maria Karasarides
Augmented reality microscopy to bridge trust between AI and pathologists
npj Precision Oncology
title Augmented reality microscopy to bridge trust between AI and pathologists
title_full Augmented reality microscopy to bridge trust between AI and pathologists
title_fullStr Augmented reality microscopy to bridge trust between AI and pathologists
title_full_unstemmed Augmented reality microscopy to bridge trust between AI and pathologists
title_short Augmented reality microscopy to bridge trust between AI and pathologists
title_sort augmented reality microscopy to bridge trust between ai and pathologists
url https://doi.org/10.1038/s41698-025-00899-5
work_keys_str_mv AT sunilbadve augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT georgelkumar augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT tobiaslang augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT elipeigin augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT jamespratt augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT robertanders augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT deyalichatterjee augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT raulsgonzalez augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT rondellpgraham augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT alyssamkrasinskas augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT xiuliliu augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT alexanderquaas augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT romilsaxena augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT namratasetia augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT lauratang augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT hanlinlwang augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT josefruschoff augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT hansulrichschildhaus augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT khaliddaifalla augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT marcpapper augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT patrickfrey augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT felixfaber augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists
AT mariakarasarides augmentedrealitymicroscopytobridgetrustbetweenaiandpathologists