A new platform for analyzing medical images
MEDiml connects feature extraction and predictive modeling through graphical and programmatic interfaces.
Researchers built MEDiml, an open-source, standards-compliant radiomics platform. It connects feature extraction, predictive modeling, and interpretation. The platform includes a Python package called MEDiml-pkg and a graphical desktop application called MEDiml-app.
Radiomics is the high-throughput extraction of quantitative features from medical images. These features include tumor shape, intensity distributions, and texture. These features can serve as candidate imaging biomarkers for cancer diagnosis, prognosis, and classification.
A typical workflow includes image preprocessing and radiomic feature extraction. It also involves feature set reduction, model training and evaluation, and interpretation using feature importance. Feature sets can be grouped into levels of radiomic complexity.
Extracting thousands of features makes it difficult to identify the most informative ones. Implementing an end-to-end workflow can also require programming expertise. This can limit accessibility for clinical researchers.
The MEDiml-app enables researchers to conduct experiments without writing code. The software can export graphically configured workflows as executable Python code. This facilitates reproducibility and collaboration between clinical researchers and computational scientists.
Key facts
- MEDiml uses both a Python package and a graphical desktop application.
- The platform was demonstrated on five oncological imaging cohorts comprising 2,104 patients.
- Feature sets are grouped into morphological, intensity-based, textural, and filter-based levels.