app.py

import os
import gradio as gr
import torch
from monai import bundle
from monai.transforms import (
    Compose,
    LoadImaged,
    EnsureChannelFirstd,
    Orientationd,
    NormalizeIntensityd,
    Activationsd,
    AsDiscreted,
    ScaleIntensityd,
)

BUNDLE_NAME = 'spleen_ct_segmentation_v0.1.0'
BUNDLE_PATH = os.path.join(torch.hub.get_dir(), 'bundle', BUNDLE_NAME)

title = "Segment Brain Tumors with MONAI!"
description = """
## Brain Tumor Segmentation 🧠
A pre-trained model for volumetric (3D) segmentation of brain tumor subregions from multimodal MRIs based on BraTS 2018 data.

## To run 🚀

Upload a image file in the format: 4 channel MRI (4 aligned MRIs T1c, T1, T2, FLAIR at 1x1x1 mm)

## Disclaimer ⚠️

This is an example, not to be used for diagnostic purposes.

## References 👀

1. Myronenko, Andriy. "3D MRI brain tumor segmentation using autoencoder regularization." International MICCAI Brainlesion Workshop. Springer, Cham, 2018. https://arxiv.org/abs/1810.11654.
2. Menze BH, et al. "The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)", IEEE Transactions on Medical Imaging 34(10), 1993-2024 (2015) DOI: 10.1109/TMI.2014.2377694
3. Bakas S, et al. "Advancing The Cancer Genome Atlas glioma MRI collections with expert segmentation labels and radiomic features", Nature Scientific Data, 4:170117 (2017) DOI:10.1038/sdata.2017.117
"""

#examples = 'examples/'

model, _, _ = bundle.load(
    name = BUNDLE_NAME,
    source = 'huggingface_hub',
    repo = 'katielink/brats_mri_segmentation_v0.1.0',
    load_ts_module=True,
)

device = "cuda:0" if torch.cuda.is_available() else "cpu"

parser = bundle.load_bundle_config(BUNDLE_PATH, 'inference.json')

preproc_transforms = Compose(
    [
        LoadImaged(keys=["image"]),
        EnsureChannelFirstd(keys="image"),
        Orientationd(keys=["image"], axcodes="RAS"),
        NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),
    ]
)
inferer = parser.get_parsed_content('inferer', lazy=True, eval_expr=True, instantiate=True)
post_transforms = Compose(
    [
        Activationsd(keys='pred', sigmoid=True),
        AsDiscreted(keys='pred', threshold=0.5),
        ScaleIntensityd(keys='image', minv=0., maxv=1.)
    ]
)

def predict(input_file, z_axis, model=model, device=device):
    data = {'image': [input_file.name]}
    data = preproc_transforms(data)
    
    model.to(device)
    model.eval()
    with torch.no_grad():
        inputs = data['image'].to(device)
        data['pred'] = inferer(inputs=inputs[None,...], network=model)
        data = post_transforms(data)
    
    input_image = data['image'].numpy()
    pred_image = data['pred'].cpu().detach().numpy()
    
    input_t1c_image = input_image[0, :, :, z_axis]
    #input_t1_image = input_image[1, :, :, z_axis]
    #input_t2_image = input_image[2, :, :, z_axis]
    #input_flair_image = input_image[3, :, :, z_axis]
    
    pred_tc_image = pred_image[0, 0, :, :, z_axis]
    #pred_et_image = pred_image[0, 1, :, :, z_axis]
    #pred_wt_image = pred_image[0, 2, :, :, z_axis]
    
    return input_t1c_image, pred_tc_image, 


iface = gr.Interface(
    fn=predict,
    inputs=[
        gr.File(label='Input file'),
        gr.Slider(0, 200, label='z-axis', value=100)
    ], 
    outputs=[
        gr.Image(label='T1C image'),
        gr.Image(label='Segmentation'),
    ],
    title=title,
    description=description,
    #examples=examples,
)

iface.launch()