FeatureForest: the power of foundation models, the usability of random forests

Abstract Analysis of biological images relies heavily on segmenting the biological objects of interest in the image before performing quantitative analysis. Deep learning (DL) is ubiquitous in such segmentation tasks, but can be cumbersome to apply, as it often requires a large amount of manual labe...

Full description

Saved in:
Bibliographic Details
Main Authors: Mehdi Seifi, Damian Dalle Nogare, Juan Manuel Battagliotti, Vera Galinova, Ananya Kedige Rao, Pierre-Henri Jouneau, Anwai Archit, AI4Life Horizon Europe Programme Consortium, Constantin Pape, Johan Decelle, Florian Jug, Joran Deschamps
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:npj Imaging
Online Access:https://doi.org/10.1038/s44303-025-00089-9
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849235031782850560
author Mehdi Seifi
Damian Dalle Nogare
Juan Manuel Battagliotti
Vera Galinova
Ananya Kedige Rao
Pierre-Henri Jouneau
Anwai Archit
AI4Life Horizon Europe Programme Consortium
Constantin Pape
Johan Decelle
Florian Jug
Joran Deschamps
author_facet Mehdi Seifi
Damian Dalle Nogare
Juan Manuel Battagliotti
Vera Galinova
Ananya Kedige Rao
Pierre-Henri Jouneau
Anwai Archit
AI4Life Horizon Europe Programme Consortium
Constantin Pape
Johan Decelle
Florian Jug
Joran Deschamps
author_sort Mehdi Seifi
collection DOAJ
description Abstract Analysis of biological images relies heavily on segmenting the biological objects of interest in the image before performing quantitative analysis. Deep learning (DL) is ubiquitous in such segmentation tasks, but can be cumbersome to apply, as it often requires a large amount of manual labeling to produce ground-truth data, and expert knowledge to train the models. More recently, large foundation models, such as SAM, have shown promising results on scientific images. They, however, require manual prompting for each object or tedious post-processing to selectively segment these objects. Here, we present FeatureForest, a method that leverages the feature embeddings of large foundation models to train a random forest classifier, thereby providing users with a rapid way of semantically segmenting complex images using only a few labeling strokes. We demonstrate the improvement in performance over a variety of datasets and provide an open-source implementation in napari that can be extended to new models.
format Article
id doaj-art-e227485bc4d747268c624fd78bd559ea
institution Kabale University
issn 2948-197X
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series npj Imaging
spelling doaj-art-e227485bc4d747268c624fd78bd559ea2025-08-20T04:02:55ZengNature Portfolionpj Imaging2948-197X2025-07-013111210.1038/s44303-025-00089-9FeatureForest: the power of foundation models, the usability of random forestsMehdi Seifi0Damian Dalle Nogare1Juan Manuel Battagliotti2Vera Galinova3Ananya Kedige Rao4Pierre-Henri Jouneau5Anwai Archit6AI4Life Horizon Europe Programme ConsortiumConstantin Pape7Johan Decelle8Florian Jug9Joran Deschamps10Computational Biology Research Center, Human TechnopoleBioimage Analysis Unit, National Facility for Data Handling and Analysis, Human TechnopoleBioimage Analysis Unit, National Facility for Data Handling and Analysis, Human TechnopoleComputational Biology Research Center, Human TechnopoleCell and Plant Physiology Laboratory, CNRS, CEA, INRAE, IRIG, Université Grenoble AlpesUniversity Grenoble Alpes, CEA, IRIG-MEMInstitute of Computer Science, University of GöttingenInstitute of Computer Science, University of GöttingenCell and Plant Physiology Laboratory, CNRS, CEA, INRAE, IRIG, Université Grenoble AlpesComputational Biology Research Center, Human TechnopoleBioimage Analysis Unit, National Facility for Data Handling and Analysis, Human TechnopoleAbstract Analysis of biological images relies heavily on segmenting the biological objects of interest in the image before performing quantitative analysis. Deep learning (DL) is ubiquitous in such segmentation tasks, but can be cumbersome to apply, as it often requires a large amount of manual labeling to produce ground-truth data, and expert knowledge to train the models. More recently, large foundation models, such as SAM, have shown promising results on scientific images. They, however, require manual prompting for each object or tedious post-processing to selectively segment these objects. Here, we present FeatureForest, a method that leverages the feature embeddings of large foundation models to train a random forest classifier, thereby providing users with a rapid way of semantically segmenting complex images using only a few labeling strokes. We demonstrate the improvement in performance over a variety of datasets and provide an open-source implementation in napari that can be extended to new models.https://doi.org/10.1038/s44303-025-00089-9
spellingShingle Mehdi Seifi
Damian Dalle Nogare
Juan Manuel Battagliotti
Vera Galinova
Ananya Kedige Rao
Pierre-Henri Jouneau
Anwai Archit
AI4Life Horizon Europe Programme Consortium
Constantin Pape
Johan Decelle
Florian Jug
Joran Deschamps
FeatureForest: the power of foundation models, the usability of random forests
npj Imaging
title FeatureForest: the power of foundation models, the usability of random forests
title_full FeatureForest: the power of foundation models, the usability of random forests
title_fullStr FeatureForest: the power of foundation models, the usability of random forests
title_full_unstemmed FeatureForest: the power of foundation models, the usability of random forests
title_short FeatureForest: the power of foundation models, the usability of random forests
title_sort featureforest the power of foundation models the usability of random forests
url https://doi.org/10.1038/s44303-025-00089-9
work_keys_str_mv AT mehdiseifi featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT damiandallenogare featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT juanmanuelbattagliotti featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT veragalinova featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT ananyakedigerao featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT pierrehenrijouneau featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT anwaiarchit featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT ai4lifehorizoneuropeprogrammeconsortium featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT constantinpape featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT johandecelle featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT florianjug featureforestthepoweroffoundationmodelstheusabilityofrandomforests
AT jorandeschamps featureforestthepoweroffoundationmodelstheusabilityofrandomforests