BiTis: Binarized Texture-based Image Synthesis for Cardiac Fibrosis

BiTis is a Python framework for generating realistic fibrosis patterns using texture-based sampling from reference histological images. It supports 2D data, provides patch-based synthesis tools, and includes metrics for structural analysis.

Features

• Synthetic fibrosis generation using reference images.
• Morphological and statistical analysis tools.

Installation

Clone and install BiTis with pip:

git clone --branch fibsim https://github.com/TiNezlobinsky/BiTis.git
cd BiTis
pip install -e .

Basic Usage

import bitis as bt
from pathlib import Path
import ast
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# Load the dataset
df = bt.datasets.tissue_dataset()
texture = df["Tissue Matrix"].iloc[1]

# Convert the texture to a binary matrix
training_tex = np.where(texture == 0, 1, 2).astype(np.float32)

# Initialize the simulation
simulation_tex = np.zeros_like(training_tex)
simulation = bt.AdaptiveSampling(simulation_tex,
                                 training_tex,
                                 max_known_pixels=30,
                                 min_known_pixels=5,
                                 max_template_size=50,
                                 min_template_size=5,
                                 n_candidates=1)

simulated_tex = simulation.run()

fig, ax = plt.subplots(1, 2, figsize=(10, 5), sharex=True, sharey=True)
ax[0].imshow(training_tex, origin='lower')
ax[1].imshow(simulated_tex, origin='lower')
plt.show()