app.py

import gradio as gr
import subprocess
import os
import shutil
from huggingface_hub import hf_hub_download
import torch
import nibabel as nib
import matplotlib as mpl
import matplotlib.pyplot as plt
import spaces  # Import spaces for GPU decoration
import numpy as np
from scipy.ndimage import center_of_mass, zoom, label, generate_binary_structure

# Define paths
MODEL_DIR = "./model"  # Local directory to store the downloaded model
DATASET_DIR = os.path.join(MODEL_DIR, "Dataset004_WML")  # Directory for Dataset004_WML
INPUT_DIR = "/tmp/input"
OUTPUT_DIR = "/tmp/output"

# Hugging Face Model Repository
REPO_ID = "FrancescoLR/FLAMeS-model"  # Replace with your actual model repository ID

import os
import subprocess

def setup_hd_bet(repo_dir="./HD-BET"):
    """
    Clones the HD-BET repository and installs it in editable mode using pip.

    Parameters:
        repo_dir (str): Directory where HD-BET will be cloned and installed.
    """
    if not os.path.exists(repo_dir):
        print("Cloning HD-BET repository...")
        subprocess.run(["git", "clone", "https://github.com/MIC-DKFZ/HD-BET", repo_dir], check=True)
    else:
        print("HD-BET repository already exists.")

    # Install the HD-BET package from source
    print("Installing HD-BET using pip...")
    subprocess.run(["pip", "install", "-e", "."], cwd=repo_dir, check=True)

# Function to download the Dataset004_WML folder
def download_model():
    if not os.path.exists(DATASET_DIR):
        os.makedirs(DATASET_DIR, exist_ok=True)
        print("Downloading Dataset004_WML.zip...")
        zip_path = hf_hub_download(repo_id=REPO_ID, filename="Dataset004_WML.zip", cache_dir=MODEL_DIR)
        subprocess.run(["unzip", "-o", zip_path, "-d", MODEL_DIR])
        print("Dataset004_WML downloaded and extracted.")
        
def resample_to_isotropic(data, affine, target_spacing=1.0):
    """
    Resamples a 3D NIfTI image to isotropic voxel size.
    
    Parameters:
        data (numpy.ndarray): The input 3D image data.
        affine (numpy.ndarray): The affine transformation matrix.
        target_spacing (float): Desired isotropic voxel spacing (in mm).
    
    Returns:
        resampled_data (numpy.ndarray): Resampled image data.
        resampled_affine (numpy.ndarray): Updated affine matrix.
    """
    # Extract current voxel dimensions from the affine matrix
    current_spacing = np.sqrt((affine[:3, :3] ** 2).sum(axis=0))

    # Compute the scaling factors for resampling
    scaling_factors = current_spacing / target_spacing

    # Resample the data using zoom
    resampled_data = zoom(data, zoom=scaling_factors, order=1)  # Linear interpolation

    # Update the affine matrix to reflect the new voxel dimensions
    resampled_affine = affine.copy()
    resampled_affine[:3, :3] /= scaling_factors[:, np.newaxis]

    return resampled_data, resampled_affine        
    
def extract_middle_slices(nifti_path, output_image_path, slice_size=180):
    """
    Extracts slices centered around the center of mass of non-zero voxels in a 3D NIfTI image.
    The slices are taken along axial, coronal, and sagittal planes and saved as a single PNG.
    """
  # Load NIfTI image
    img = nib.load(nifti_path)
    data = img.get_fdata()
    affine = img.affine

    # Resample the image to 1 mm isotropic
    resampled_data, _ = resample_to_isotropic(data, affine, target_spacing=1.0)

    # Compute the center of mass of non-zero voxels
    com = center_of_mass(resampled_data > 0)
    center = np.round(com).astype(int)

    # Define half the slice size
    half_size = slice_size // 2

def extract_middle_slices(nifti_path, output_image_path, slice_size=180, center=None, label_components=False):
    """
    Extracts slices from a 3D NIfTI image. 
    If label_components=True, it assigns different labels (colors) to each connected component (26-connectivity)
    and returns the labeled 3D mask.
    
    Returns:
        labeled_data (np.ndarray): The 3D array (either labeled or original).
    """
    # Load NIfTI image
    img = nib.load(nifti_path)
    data = img.get_fdata()
    affine = img.affine

    # Resample the image to 1 mm isotropic
    resampled_data, _ = resample_to_isotropic(data, affine, target_spacing=1.0)

    # Optionally label connected components
    if label_components:
        structure = generate_binary_structure(3, 3)  # 3D, 26-connectivity
        labeled_data, num_features = label(data > 0, structure=structure)
        labeled_data_resampled, num_features = label(resampled_data > 0, structure=structure)
    else:
        labeled_data = resampled_data
        num_features = None  # Not needed if we're not labeling
        labeled_data_resampled = resampled_data

    # Compute or reuse the center of mass
    if center is None:
        com = center_of_mass(labeled_data_resampled > 0)
        center = np.round(com).astype(int)

    # Define half the slice size
    half_size = slice_size // 2

    # Function to extract and pad slices
    def extract_2d_slice(data, center, axis):
        slices = [slice(None)] * 3
        slices[axis] = center[axis]
        extracted_slice = data[tuple(slices)]

        remaining_axes = [i for i in range(3) if i != axis]
        cropped_slice = extracted_slice[
            max(center[remaining_axes[0]] - half_size, 0):min(center[remaining_axes[0]] + half_size, extracted_slice.shape[0]),
            max(center[remaining_axes[1]] - half_size, 0):min(center[remaining_axes[1]] + half_size, extracted_slice.shape[1]),
        ]

        pad_height = slice_size - cropped_slice.shape[0]
        pad_width = slice_size - cropped_slice.shape[1]
        padded_slice = np.pad(cropped_slice,
                              ((pad_height // 2, pad_height - pad_height // 2),
                               (pad_width // 2, pad_width - pad_width // 2)),
                              mode='constant', constant_values=0)
        return padded_slice

    # Extract slices
    axial_slice = extract_2d_slice(labeled_data_resampled, center, axis=2)
    coronal_slice = extract_2d_slice(labeled_data_resampled, center, axis=1)
    sagittal_slice = extract_2d_slice(labeled_data_resampled, center, axis=0)

    # Apply rotations
    axial_slice = np.rot90(axial_slice, k=-1)
    coronal_slice = np.rot90(coronal_slice, k=1)
    coronal_slice = np.rot90(coronal_slice, k=2)
    sagittal_slice = np.rot90(sagittal_slice, k=1)
    sagittal_slice = np.rot90(sagittal_slice, k=2)

    # Create subplots
    fig, axes = plt.subplots(1, 3, figsize=(12, 4))

    # Choose colormap
    if label_components:
        # Create 256 pastel colors
        pastel = plt.cm.Pastel1(np.linspace(0, 1, 256))
        np.random.seed(42)  # For reproducibility
        shuffled_colors = pastel[1:].copy()
        np.random.shuffle(shuffled_colors)
        final_colors = np.vstack([np.array([0, 0, 0, 1]), shuffled_colors])
 
        custom_cmap = mpl.colors.ListedColormap(final_colors)
        cmap = custom_cmap  # Colorful
        vmin = 0
        vmax = num_features
    else:
        cmap = "gray"  # Normal
        vmin = None
        vmax = None

    # Plot slices
    for idx, slice_data in enumerate([axial_slice, coronal_slice, sagittal_slice]):
        ax = axes[idx]
        im = ax.imshow(slice_data, cmap=cmap, origin="lower", vmin=vmin, vmax=vmax)
        ax.axis("off")

    # Save figure
    plt.tight_layout()
    plt.savefig(output_image_path, bbox_inches="tight", pad_inches=0)
    plt.close()

    # Return the labeled mask
    return labeled_data
    
# Function to run nnUNet inference
@spaces.GPU(duration=90)  # Decorate the function to allocate GPU for its execution
def run_nnunet_predict(nifti_file,hd_bet=False):
    # Prepare directories
    os.makedirs(INPUT_DIR, exist_ok=True)
    os.makedirs(OUTPUT_DIR, exist_ok=True)

    # Extract the original filename without the extension
    original_filename = os.path.basename(nifti_file.name)
    base_filename = original_filename.replace(".nii.gz", "")
    
    # Save the uploaded file to the input directory
    input_path = os.path.join(INPUT_DIR, "image_0000.nii.gz")
    os.rename(nifti_file.name, input_path)  # Move the uploaded file to the expected input location

    if hd_bet:
       # Apply skull-stripping with HD-BET
       hd_bet_output_path = os.path.join(INPUT_DIR, "image_0000.nii.gz")
       try:
          subprocess.run([
          "hd-bet",
          "-i", input_path,
          "-o", hd_bet_output_path,
          "-device", "cuda",  # or "cpu"
          "--disable_tta" ], check=True)
          print("Skull-stripping completed.")
          input_path = hd_bet_output_path
       except subprocess.CalledProcessError as e:
          return f"HD-BET Error: {e}"
    
    # Debugging: List files in the /tmp/input directory
    print("Files in /tmp/input:")
    print(os.listdir(INPUT_DIR))

    # Set environment variables for nnUNet
    os.environ["nnUNet_results"] = MODEL_DIR

    # Construct and run the nnUNetv2_predict command
    command = [
        "nnUNetv2_predict",
        "-i", INPUT_DIR,
        "-o", OUTPUT_DIR,
        "-d", "004",                  # Dataset ID
        "-c", "3d_fullres",           # Configuration
        "-tr", "nnUNetTrainer_8000epochs",
        "-device", "cuda",  # Explicitly use GPU
    ]
    print("Files in /tmp/output:")
    print(os.listdir(OUTPUT_DIR))
    try:
        subprocess.run(command, check=True)

        # Rename the output file to match the original input filename
        output_file = os.path.join(OUTPUT_DIR, "image.nii.gz")
        new_output_file = os.path.join(OUTPUT_DIR, f"{base_filename}_LesionMask.nii.gz")
        if os.path.exists(output_file):
            os.rename(output_file, new_output_file)

            # Compute center of mass for the input image
            img = nib.load(input_path)
            data = img.get_fdata()
            affine = img.affine
            resampled_data, _ = resample_to_isotropic(data, affine, target_spacing=1.0)
            com = center_of_mass(resampled_data > 0)  # Center of mass
            center = np.round(com).astype(int)        # Round to integer

            # Extract and save 2D slices
            input_slice_path = os.path.join(OUTPUT_DIR, f"{base_filename}_input_slice.png")
            output_slice_path = os.path.join(OUTPUT_DIR, f"{base_filename}_output_slice.png")
            image = extract_middle_slices(input_path, input_slice_path, center=center)
            labeled_mask = extract_middle_slices(new_output_file, output_slice_path, center=center, label_components=True)

            # Load the binary lesion mask to get its affine
            output_img = nib.load(new_output_file)

            labeled_mask_path = os.path.join(OUTPUT_DIR, f"{base_filename}_LabeledClusters.nii.gz")
            nib.save(nib.Nifti1Image(labeled_mask.astype(np.int16), output_img.affine), labeled_mask_path)

            # Return paths for the Gradio interface
            return new_output_file, input_slice_path, output_slice_path, labeled_mask_path
        else:
            return "Error: Output file not found."
    except subprocess.CalledProcessError as e:
        return f"Error: {e}"

# Gradio interface with adjusted layout
with gr.Blocks() as demo:
    gr.Markdown("""
   # 🔥 FLAMeS: FLAIR Lesion Segmentation for Multiple Sclerosis

    Upload a FLAIR brain MRI in NIfTI format (.nii.gz) to generate a binary segmentation of multiple sclerosis lesions.  
    FLAMeS is based on the nnUNet framework<sup>2</sup> and was trained on 668 MRI scans acquired using Siemens, GE, and Philips 1.5T and 3T scanners<sup>1</sup>.
    We suggest skull-stripping the image in advance using [SynthStrip](https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/) with the `--no-csf` flag for optimal results. If that's not feasible, you can still upload your image as-is and enable the "Apply skull-stripping" option below.

    Inference takes approximately 1 minute per MRI, with processing limited to one scan at a time due to Hugging Face's zero-GPU usage constraints. To process multiple cases simultaneously, install the [nnUNet v2](https://github.com/MIC-DKFZ/nnUNet), download [FLAMeS's model](https://huggingface.co/FrancescoLR/FLAMeS-model) and run it locally using your own GPU or CPU setup.
    
    **Disclaimer:** Uploaded data is stored temporarily, no one has access to it, and it is deleted when the app is closed. For details, see [Gradio's file access guide](https://www.gradio.app/main/guides/file-access). Human subjects data should only be uploaded for processing if permitted by your institution's human subjects protection office.
    This is a research tool and is not intended for clinical use. Clinical decisions should not be based on the outputs of this tool.
    
    """)

    with gr.Row():
        with gr.Column(scale=1):
            flair_input = gr.File(label="Upload a FLAIR Image (.nii.gz)")
            hd_bet = gr.Checkbox(label="Apply skull-stripping", value=False)
            submit_button = gr.Button("Submit")
        with gr.Column(scale=2):
            seg_output = gr.File(label="Download the Lesion Segmentation Mask")
            clusters_output = gr.File(label="Download the Labeled Lesion Segmentation Mask")
            input_img = gr.Image(label="Input: FLAIR image")
            output_img = gr.Image(label="Output: Binary Lesion Mask")
            

    gr.Markdown("""
    **If you find this tool useful, please consider citing:**

    1. FLAMeS: A Robust Deep Learning Model for Automated Multiple Sclerosis Lesion Segmentation  
   Dereskewicz, E., La Rosa, F., dos Santos Silva, J., Sizer, E., Kohli, A., Wynen, M., ... & Beck, E. S.
   *medRxiv (2025)  
   DOI: [10.1177/13524585231169437](https://doi.org/10.1101/2025.05.19.25327707)

    2. nnU-Net: A Self-Configuring Method for Deep Learning-Based Biomedical Image Segmentation  
   F. Isensee, P. F. Jaeger, S. A. Kohl, J. Petersen, & K. H. Maier-Hein.  
   *Nature Methods.* 2021;18(2):203-211.  
   DOI: [10.1038/s41592-020-01008-z](https://www.nature.com/articles/s41592-020-01008-z)
    """)

    submit_button.click(
        fn=run_nnunet_predict,
        inputs=[flair_input, hd_bet],
        outputs=[seg_output, input_img, output_img, clusters_output]
    )

# Debugging GPU environment
if torch.cuda.is_available():
    print(f"GPU is available: {torch.cuda.get_device_name(0)}")
else:
    print("No GPU available. Falling back to CPU.")
    os.system("nvidia-smi")

setup_hd_bet()
download_model()

if __name__ == "__main__":
    demo.launch(share=True)