neuroflux 2.0

MRI Brain Tumor Diagnosis and Grad-CAM Visualization

Neuroflux

Neuroflux is a Python package that analyzes MRI scans of brains and highlights regions of tumor damage on a heat map.

Installation

To install neuroflux, you can use pip:

pip install neuroflux

The recommended importation is:

from neuroflux import (
    prepare_model,
    display_slice,
    display_grid,
)

Setup

To begin, declare the following variables:

• case_folder: The folder where the input MRI scans are located (set it to "" if the input files are in the same folder as the program)
• input_flair: The file name of the input FLAIR
• input_t1ce: The file name of the input T1ce
• model_weights: The file name of the predefined model weights

Example setup:

case_folder = "BraTS20"
input_flair = "01_flair.nii"
input_t1ce = "01_t1ce.nii"
model_weights = "model_.weights.h5"

Usage

Neuroflux has three main functions:

• prepare_model(model_weights, img_size=128)
  • Creates the U-Net (first step)
• display_slice(folder, input_flair, input_t1ce, slice_num, model, target_layer_name="conv2d_1", img_size=128)
  • Displays the Grad-CAM heat map at a given slice number
• display_grid(folder, input_flair, input_t1ce, model, num_slices=84, rows=7, cols=12, start_slice=20, target_layer_name="conv2d_1")
  • Displays a grid containing a given number of slices beginning at a given slice number

Example usage:

model = prepare_model(model_weights=model_weights, img_size=img_size)
display_slice(folder=case_folder, input_flair=input_flair, input_t1ce=input_t1ce, slice_num=77, model=model)
display_grid(folder=case_folder,input_flair=input_flair, input_t1ce=input_t1ce, model=model)

Notes

• The model weights are not built into this package. Our GitHub contains two options to attain model weights: Use our pre-trained weights (recommend) or run the Google Colab notebook to create your own unique weights
• To get more involved, see our website, read our research paper, or contact us at [e-mail].