Open-source Phikon Histopathology Model - Free to Extract Features from Histological Image Patches

Phikon

Developed by owkin

Phikon is a self-supervised learning model for histopathology based on iBOT training, primarily used for extracting features from histology image patches.

Image Classification

Transformers

EnglishOpen Source License:Other #Histopathological feature extraction #Cancer subtype classification #Self-supervised learning

Downloads 741.63k

Release Time : 9/21/2023

Model Overview

Phikon is a Vision Transformer model for histopathological image analysis, extracting features from histology image patches through self-supervised learning, suitable for various cancer subtype classification tasks.

Model Features

Self-supervised learning

Uses the iBOT framework for self-supervised pre-training, learning image features without labeled data.

Large-scale pre-training

Pre-trained on 40 million pan-cancer image patches from the TGCA dataset.

Histopathology-specific

Optimized specifically for histopathological images, suitable for cancer subtype analysis.

Model Capabilities

Histopathological image feature extraction

Cancer subtype classification

Image patch analysis

Use Cases

Medical research

Cancer subtype classification

Classify histopathological images to identify different cancer subtypes.

Biomarker discovery

Extract features from histology images for biomarker research.

Clinical diagnosis

Auxiliary diagnosis

Provide auxiliary diagnostic information for pathologists.

🚀 Model Card for Phikon

Phikon is a self - supervised learning model for histopathology trained with iBOT. It can be used for feature extraction from histology image tiles and cancer classification on various cancer subtypes.

🚀 Quick Start

🎉 Important Note Check out the latest version of Phikon here: [Phikon - v2](https://huggingface.co/owkin/phikon - v2)

To learn more about how to use the model, we encourage you to read our blog post and view this Colab notebook.

✨ Features

Feature Extraction: The primary use of the Phikon model is for feature extraction from histology image tiles.
Cancer Classification: It can be used for cancer classification on a variety of cancer subtypes and can also be finetuned to specialise on cancer subtypes.

📚 Documentation

Model Description

Property	Details
Developed by	Owkin
Funded by	Owkin and IDRIS
Model Type	Vision Transformer Base
Model Stats	Params (M): 85.8; Image size: 224 x 224 x 3
Paper	Scaling Self - Supervised Learning for Histopathology with Masked Image Modeling. A. Filiot et al., medRxiv 2023.07.21.23292757; doi: https://doi.org/10.1101/2023.07.21.23292757
Pretrain Dataset	40 million pan - cancer tiles extracted from TGCA
Original	https://github.com/owkin/HistoSSLscaling/
License	Owkin non - commercial license

Uses

Direct Use

The primary use of the Phikon model is for feature extraction from histology image tiles.

Downstream Use

The model can be used for cancer classification on a variety of cancer subtypes and can also be finetuned to specialise on cancer subtypes.

🔧 Technical Details

Compute Infrastructure

All the models we built were trained on the French Jean Zay cluster.

Hardware

NVIDIA V100 GPUs with 32Gb RAM

Software

PyTorch 1.13.1

BibTeX entry and citation info

@article{Filiot2023ScalingSSLforHistoWithMIM,
	author       = {Alexandre Filiot and Ridouane Ghermi and Antoine Olivier and Paul Jacob and Lucas Fidon and Alice Mac Kain and Charlie Saillard and Jean - Baptiste Schiratti},
	title        = {Scaling Self - Supervised Learning for Histopathology with Masked Image Modeling},
	elocation - id = {2023.07.21.23292757},
	year         = {2023},
	doi          = {10.1101/2023.07.21.23292757},
	publisher    = {Cold Spring Harbor Laboratory Press},
	url          = {https://www.medrxiv.org/content/early/2023/07/26/2023.07.21.23292757},
	eprint       = {https://www.medrxiv.org/content/early/2023/07/26/2023.07.21.23292757.full.pdf},
	journal      = {medRxiv}
}

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご