Automating liver biopsy segmentation with a robust, open-source tool for pathology research: the HOTSPoT model

Giorgio Cazzaniga; Vincenzo L'Imperio; Emanuela Bonoldi; Maria-Carlota Londono; Joao Madaleno; Augusta Cipriano; Tom J. G. Gevers; Iryna V. Samarska; Ozgur M. Koc; Alejandra Villamil; Maria Florencia Sanchez; Vincenza Calvaruso; Alberto Quattrocchi; Daniela Cabibi; Bastian Engel; Federica Malinverno; Elisa Merelli; Laura Cristoferi; Marco Carbone; Fabio Pagni; Pietro Invernizzi; Alessio Gerussi

doi:10.1038/s41746-025-01870-1

Automating liver biopsy segmentation with a robust, open-source tool for pathology research: the HOTSPoT model

Giorgio Cazzaniga, Vincenzo L'Imperio, Emanuela Bonoldi, Maria-Carlota Londono, Joao Madaleno, Augusta Cipriano, Tom J. G. Gevers, Iryna V. Samarska, Ozgur M. Koc, Alejandra Villamil, Maria Florencia Sanchez, Vincenza Calvaruso, Alberto Quattrocchi, Daniela Cabibi, Bastian Engel, Federica Malinverno, Elisa Merelli, Laura Cristoferi, Marco Carbone, Fabio PagniPietro Invernizzi, Alessio Gerussi^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › Academic › peer-review

Abstract

Artificial intelligence applications in liver pathology remain limited, with existing tools either narrowly focused or lacking external validation. This study introduces HOTSPoT, an open-source, validated transformer-based model for automated segmentation of portal tracts in H&E-stained liver biopsy whole slide images. A multi-institutional dataset of 223 cases was used, with annotations by expert hepatopathologists. HOTSPoT achieved high performance with mean Dice scores of 0.92 (train/val) and 0.91 (test), and mean IoUs of 0.86, 0.85, and 0.84, respectively, showing minimal domain shift. Automated portal tract quantification showed strong concordance with manual assessments (κ up to 0.90), and portal area correlated with fibrosis stage (r = 0.87, p < 0.001). The model is available as a TorchScript file with a modified WSInfer library, enabling efficient WSI-level inference and integration with QuPath for advanced pathology analysis.

Original language	English
Article number	455
Number of pages	9
Journal	npj Digital Medicine
Volume	8
Issue number	1
DOIs	https://doi.org/10.1038/s41746-025-01870-1
Publication status	Published - Dec 2025

Access to Document

10.1038/s41746-025-01870-1Licence: CC BY-NC-ND

Cite this

Cazzaniga, G., L'Imperio, V., Bonoldi, E., Londono, M.-C., Madaleno, J., Cipriano, A., Gevers, T. J. G., Samarska, I. V., Koc, O. M., Villamil, A., Sanchez, M. F., Calvaruso, V., Quattrocchi, A., Cabibi, D., Engel, B., Malinverno, F., Merelli, E., Cristoferi, L., Carbone, M., ... Gerussi, A. (2025). Automating liver biopsy segmentation with a robust, open-source tool for pathology research: the HOTSPoT model. npj Digital Medicine, 8(1), Article 455. https://doi.org/10.1038/s41746-025-01870-1

@article{205c57894859420aa32d5e5a27bc454b,

title = "Automating liver biopsy segmentation with a robust, open-source tool for pathology research: the HOTSPoT model",

abstract = "Artificial intelligence applications in liver pathology remain limited, with existing tools either narrowly focused or lacking external validation. This study introduces HOTSPoT, an open-source, validated transformer-based model for automated segmentation of portal tracts in H\&E-stained liver biopsy whole slide images. A multi-institutional dataset of 223 cases was used, with annotations by expert hepatopathologists. HOTSPoT achieved high performance with mean Dice scores of 0.92 (train/val) and 0.91 (test), and mean IoUs of 0.86, 0.85, and 0.84, respectively, showing minimal domain shift. Automated portal tract quantification showed strong concordance with manual assessments (κ up to 0.90), and portal area correlated with fibrosis stage (r = 0.87, p < 0.001). The model is available as a TorchScript file with a modified WSInfer library, enabling efficient WSI-level inference and integration with QuPath for advanced pathology analysis.",

author = "Giorgio Cazzaniga and Vincenzo L'Imperio and Emanuela Bonoldi and Maria-Carlota Londono and Joao Madaleno and Augusta Cipriano and Gevers, \{Tom J. G.\} and Samarska, \{Iryna V.\} and Koc, \{Ozgur M.\} and Alejandra Villamil and Sanchez, \{Maria Florencia\} and Vincenza Calvaruso and Alberto Quattrocchi and Daniela Cabibi and Bastian Engel and Federica Malinverno and Elisa Merelli and Laura Cristoferi and Marco Carbone and Fabio Pagni and Pietro Invernizzi and Alessio Gerussi",

year = "2025",

month = dec,

doi = "10.1038/s41746-025-01870-1",

language = "English",

volume = "8",

journal = "npj Digital Medicine",

issn = "2398-6352",

publisher = "Springer Nature",

number = "1",

}

Cazzaniga, G, L'Imperio, V, Bonoldi, E, Londono, M-C, Madaleno, J, Cipriano, A, Gevers, TJG, Samarska, IV , Koc, OM, Villamil, A, Sanchez, MF, Calvaruso, V, Quattrocchi, A, Cabibi, D, Engel, B, Malinverno, F, Merelli, E, Cristoferi, L, Carbone, M, Pagni, F, Invernizzi, P & Gerussi, A 2025, 'Automating liver biopsy segmentation with a robust, open-source tool for pathology research: the HOTSPoT model', npj Digital Medicine, vol. 8, no. 1, 455. https://doi.org/10.1038/s41746-025-01870-1

TY - JOUR

T1 - Automating liver biopsy segmentation with a robust, open-source tool for pathology research

T2 - the HOTSPoT model

AU - Cazzaniga, Giorgio

AU - L'Imperio, Vincenzo

AU - Bonoldi, Emanuela

AU - Londono, Maria-Carlota

AU - Madaleno, Joao

AU - Cipriano, Augusta

AU - Gevers, Tom J. G.

AU - Samarska, Iryna V.

AU - Koc, Ozgur M.

AU - Villamil, Alejandra

AU - Sanchez, Maria Florencia

AU - Calvaruso, Vincenza

AU - Quattrocchi, Alberto

AU - Cabibi, Daniela

AU - Engel, Bastian

AU - Malinverno, Federica

AU - Merelli, Elisa

AU - Cristoferi, Laura

AU - Carbone, Marco

AU - Pagni, Fabio

AU - Invernizzi, Pietro

AU - Gerussi, Alessio

PY - 2025/12

Y1 - 2025/12

N2 - Artificial intelligence applications in liver pathology remain limited, with existing tools either narrowly focused or lacking external validation. This study introduces HOTSPoT, an open-source, validated transformer-based model for automated segmentation of portal tracts in H&E-stained liver biopsy whole slide images. A multi-institutional dataset of 223 cases was used, with annotations by expert hepatopathologists. HOTSPoT achieved high performance with mean Dice scores of 0.92 (train/val) and 0.91 (test), and mean IoUs of 0.86, 0.85, and 0.84, respectively, showing minimal domain shift. Automated portal tract quantification showed strong concordance with manual assessments (κ up to 0.90), and portal area correlated with fibrosis stage (r = 0.87, p < 0.001). The model is available as a TorchScript file with a modified WSInfer library, enabling efficient WSI-level inference and integration with QuPath for advanced pathology analysis.

AB - Artificial intelligence applications in liver pathology remain limited, with existing tools either narrowly focused or lacking external validation. This study introduces HOTSPoT, an open-source, validated transformer-based model for automated segmentation of portal tracts in H&E-stained liver biopsy whole slide images. A multi-institutional dataset of 223 cases was used, with annotations by expert hepatopathologists. HOTSPoT achieved high performance with mean Dice scores of 0.92 (train/val) and 0.91 (test), and mean IoUs of 0.86, 0.85, and 0.84, respectively, showing minimal domain shift. Automated portal tract quantification showed strong concordance with manual assessments (κ up to 0.90), and portal area correlated with fibrosis stage (r = 0.87, p < 0.001). The model is available as a TorchScript file with a modified WSInfer library, enabling efficient WSI-level inference and integration with QuPath for advanced pathology analysis.

U2 - 10.1038/s41746-025-01870-1

DO - 10.1038/s41746-025-01870-1

M3 - Article

SN - 2398-6352

VL - 8

JO - npj Digital Medicine

JF - npj Digital Medicine

IS - 1

M1 - 455

ER -

Automating liver biopsy segmentation with a robust, open-source tool for pathology research: the HOTSPoT model

Abstract

Access to Document

Fingerprint

Cite this