Semantic Scholar Open Access 2025 14 sitasi

Foundation Models -- A Panacea for Artificial Intelligence in Pathology?

N. Mulliqi A. Blilie X. Ji K. Szolnoky Henrik Olsson +100 lainnya

Lihat Sumber

Abstrak

The role of artificial intelligence (AI) in pathology has evolved from aiding diagnostics to uncovering predictive morphological patterns in whole slide images (WSIs). Recently, foundation models (FMs) leveraging self-supervised pre-training have been widely advocated as a universal solution for diverse downstream tasks. However, open questions remain about their clinical applicability and generalization advantages over end-to-end learning using task-specific (TS) models. Here, we focused on AI with clinical-grade performance for prostate cancer diagnosis and Gleason grading. We present the largest validation of AI for this task, using over 100,000 core needle biopsies from 7,342 patients across 15 sites in 11 countries. We compared two FMs with a fully end-to-end TS model in a multiple instance learning framework. Our findings challenge assumptions that FMs universally outperform TS models. While FMs demonstrated utility in data-scarce scenarios, their performance converged with - and was in some cases surpassed by - TS models when sufficient labeled training data were available. Notably, extensive task-specific training markedly reduced clinically significant misgrading, misdiagnosis of challenging morphologies, and variability across different WSI scanners. Additionally, FMs used up to 35 times more energy than the TS model, raising concerns about their sustainability. Our results underscore that while FMs offer clear advantages for rapid prototyping and research, their role as a universal solution for clinically applicable medical AI remains uncertain. For high-stakes clinical applications, rigorous validation and consideration of task-specific training remain critically important. We advocate for integrating the strengths of FMs and end-to-end learning to achieve robust and resource-efficient AI pathology solutions fit for clinical use.

Topik & Kata Kunci

Computer Science

Penulis (105)

N. Mulliqi

A. Blilie

X. Ji

K. Szolnoky

Henrik Olsson

S. E. Boman

Matteo Titus

G. Gonzalez

Julia Anna Mielcarz

Masi Valkonen

E. Gudlaugsson

S. R. Kjosavik

José Asenjo

Marcello Gambacorta

Paolo Libretti

M. Braun

R. Kordek

Roman Lowicki

K. Hotakainen

Paivi Vare

B. Pedersen

Karina Dalsgaard Sørensen

B. Ulhøi

P. Ruusuvuori

Brett Delahunt

H. Samaratunga

Toyonori Tsuzuki

E. Janssen

L. Egevad

Martin Eklund

Kimmo Kartasalo Department of Medical Epidemiology

Biostatistics

Karolinska Institutet

Stockholm

Sweden

D. '. Pathology

S. Hospital

Stavanger

Norway.

F. O. Sciences

University of Stavanger

Department of Preventive Medicine

Surgery

Institute of Biomedicine

U. Turku

Turku

Finland.

The General Practice

Care Coordination Research Group

D. Health

Primary Care

Faculty of Veterinary Medicine

Universityof Bergen

Synlab

Madrid

Spain.

Brescia

Italy

Chair of Oncology

Medical University of Lodz

Lodz

Poland

1st Department of Urology

D. Chemistry

Hematology

U. Helsinki

Helsinki

Laboratory Services

Mehilainen Oy

Mehilainen Lansi-Pohja Hospital

Kemi

D. Radiology

Aarhus University Hospital

Aarhus

Denmark.

Department of Preventive Medicine

A. University

InFLAMES Research Flagship

H. Technology

Tampere University

Tampere

M. I. O. M. Research

Wellington

N. Zealand.

Department of Radiation Oncology

Pathology

Aquesta Uropathology

U. Queensland

Qld

Brisbane

Australia

D. '. Pathology

School of Clinical Medicine

Aichi Medical University

Nagoya

Japan

D. Chemistry

Bioscience

Environmental Engineering

Institute for Biomedicine

Glycomics

Griffith University

Queensland

Department of Medical Epidemiology

SciLifeLab

Format Sitasi

APA MLA BibTeX

Mulliqi, N., Blilie, A., Ji, X., Szolnoky, K., Olsson, H., Boman, S.E. et al. (2025). Foundation Models -- A Panacea for Artificial Intelligence in Pathology?. https://www.semanticscholar.org/paper/516d76eb4547667ff9191853afc407fedb0564e4

@article{mulliqi2025, author = {N. Mulliqi and A. Blilie and X. Ji and K. Szolnoky and Henrik Olsson and S. E. Boman and Matteo Titus and G. Gonzalez and Julia Anna Mielcarz and Masi Valkonen and E. Gudlaugsson and S. R. Kjosavik and José Asenjo and Marcello Gambacorta and Paolo Libretti and M. Braun and R. Kordek and Roman Lowicki and K. Hotakainen and Paivi Vare and B. Pedersen and Karina Dalsgaard Sørensen and B. Ulhøi and P. Ruusuvuori and Brett Delahunt and H. Samaratunga and Toyonori Tsuzuki and E. Janssen and L. Egevad and Martin Eklund and Kimmo Kartasalo Department of Medical Epidemiology and Biostatistics and Karolinska Institutet and Stockholm and Sweden and D. '. Pathology and S. Hospital and Stavanger and Norway. and F. O. Sciences and University of Stavanger and Department of Preventive Medicine and Surgery and Institute of Biomedicine and U. Turku and Turku and Finland. and The General Practice and Care Coordination Research Group and D. Health and Primary Care and Faculty of Veterinary Medicine and Universityof Bergen and Synlab and Madrid and Spain. and Brescia and Italy and Chair of Oncology and Medical University of Lodz and Lodz and Poland and 1st Department of Urology and D. Chemistry and Hematology and U. Helsinki and Helsinki and Laboratory Services and Mehilainen Oy and Mehilainen Lansi-Pohja Hospital and Kemi and D. Radiology and Aarhus University Hospital and Aarhus and Denmark. and Department of Preventive Medicine and A. University and InFLAMES Research Flagship and H. Technology and Tampere University and Tampere and M. I. O. M. Research and Wellington and N. Zealand. and Department of Radiation Oncology and Pathology and Aquesta Uropathology and U. Queensland and Qld and Brisbane and Australia and D. '. Pathology and School of Clinical Medicine and Aichi Medical University and Nagoya and Japan and D. Chemistry and Bioscience and Environmental Engineering and Institute for Biomedicine and Glycomics and Griffith University and Queensland and Department of Medical Epidemiology and SciLifeLab}, title = {Foundation Models -- A Panacea for Artificial Intelligence in Pathology?}, year = {2025}, journal = {Semantic Scholar}, url = {https://www.semanticscholar.org/paper/516d76eb4547667ff9191853afc407fedb0564e4}, }

Akses Cepat

PDF tidak tersedia langsung

Cek di sumber asli →

Lihat di Sumber

Informasi Jurnal

Tahun Terbit: 2025
Bahasa: en
Total Sitasi: 14×
Sumber Database: Semantic Scholar
Akses: Open Access ✓