OncoTUM models

    This repository hosts pretrained neural network models for OncoTUM, 
    a key software package within the umbrella project Onco* for modelling and numerical simulations of tumours. OncoTUM is designed to facilitate tumour segmentations from medical images, leveraging state-of-the-art deep learning techniques.


Purpose

    The pretrained models in this repository are required for using the inference function of OncoTUM. These models have been trained on relevant datasets (BraTS 2020) to ensure 
    high accuracy and performance in tumor segmentation and its classification.


Usage

    To utilise the inference functionality of OncoTUM, download the appropriate pretrained models from this repository and ensure they are correctly linked to the OncoTUM software package. 
    Detailed instructions for setup and integration can be found in the 
    OncoTUM documentation.


Content

In order to remain with most possible flexibility, the modality agnostic implementation allows to perform segmentation with a subset of the gold standard modalities.

Brain tumour segmentation

    Full modal model: trained to all gold standard modalities (t1, t1gd, t2, flair).
    Single modality model: trained to single modalities of the gold standard.
    Null image: Empty image for training.