The medical imaging analysis using machine learning encounters challenges such as limited and imbalanced datasets, often constrained by privacy concerns related to patient information. The Medicraft project addresses these limitations by developing software capable of generating synthetic and diverse medical datasets from imaging information. In cooperation with the University Clinical Hospital in Poznan this tool utilizes OCT eye scans, featuring images with abnormalities like tumors and melanomas. To reduce the scarcity of real data, for medically rare cases, the solution uses diffusion models (Denoising Diffusion Probabilistic Model) to create synthetic balanced datasets which can facilitate better research and education in machine learning for medical image analysis and classification. The project uses a data-driven learning approach focused on analyzing the generated data, where synthetic images are analyzed for accuracy with the medical team, in order to achieve the best possible result.
The documentation for the project can be found here.
Create an environment from the environment.yml
file:
conda env create -f environment.yml
Install the required packages using pip:
pip install -r requirements.txt
To install the project as a package, run the following command:
pip install medicraft
After successfully installing Medicraft, it is highly recommended to familiarize yourself with the documentation. You can find the detailed documentation here.
To learn how to run the project, refer to the usage instructions provided in the documentation here.
For running a project create config.yml
file and run the following command:
python src/main.py -f config.yml
See config examples here: configs
For more information on how to install the project, please refer to the installation guide.