radiomics-suite-json9112 0.1.1

Advanced radiomics analysis suite for raw segmented imaging and clinical outcome modeling

radiomics-suite-json9112

radiomics-suite-json9112 wraps advanced radiomics analysis behind a single high-level API.

It accepts raw segmented image folders plus a clinical table, builds traditional and deep radiomics feature spaces internally, synthesizes temporary latent embeddings during analysis, and writes only the final result artifacts by default.

Install

pip install radiomics-suite-json9112

Example

from radiomics_suite import EnhancedRadiomicsPipeline

pipeline = EnhancedRadiomicsPipeline()
results = pipeline.run_adv_analysis(
    segmented_folder="/path/to/Segmented images",
    clinical_file="/path/to/master_dataset_normalized_immunotherapy.csv",
    outcome_columns=["OS_12m", "OS_36m", "PFS_12m", "PFS_36m", "DCB", "PNEUMONITIS"],
    light=False,
)

Proteomics-style aliasing is also supported for common outcome requests:

from radiomics_suite import EnhancedRadiomicsPipeline

pipeline = EnhancedRadiomicsPipeline()
results = pipeline.run_adv_analysis(
    signal_folder="/path/to/Segmented images",
    clinical_file="/path/to/master_dataset_normalized_immunotherapy.csv",
    outcome_columns=["PNEUMONITIS", "grade >=3"],
    light=True,
)

Precomputed features can be reused when needed:

from radiomics_suite import EnhancedRadiomicsPipeline

pipeline = EnhancedRadiomicsPipeline()
results = pipeline.run_adv_analysis(
    clinical_file="/path/to/master_dataset_normalized_immunotherapy.csv",
    traditional_features_file="/path/to/tumor_radiomics_features.csv",
    deep_features_file="/path/to/complete_cool_unet_radiomics.csv",
    outcome_columns=["DCB"],
    light=True,
)

Input Layout

The segmented folder should contain patient subfolders such as L30 segmented/, each with paired files like:

• image_00042.png
• image_00042_mask.png

The clinical table may be .csv, .xlsx, or .xls.

Notes

• Raw segmented images are the preferred interface.
• Intermediate embeddings are kept in-memory unless persist_intermediates=True.
• light=True runs a smaller model/search configuration.
• Returned results include output_dir, feature_generation, and per-outcome best-scenario summaries.