{
    "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
    "version": "0.0.1",
    "changelog": {
        "0.0.1": "Initial version"
    },
    "monai_version": "1.5.0",
    "pytorch_version": "2.6.0",
    "numpy_version": "1.26.4",
    "optional_packages_version": {},
    "required_packages_version": {
        "setuptools": "75.8.0",
        "opencv-python-headless": "4.11.0.86",
        "pandas": "2.3.0",
        "seaborn": "0.13.2",
        "scikit-learn": "1.6.1",
        "progressbar": "2.5",
        "pydicom": "3.0.1",
        "fire": "0.7.0",
        "torchvision": "0.21.0",
        "detectron2": "0.6",
        "lxml": "5.4.0",
        "pillow": "11.2.1"
    },
    "name": "retinalOCT_RPD_segmentation",
    "task": "Reticular Pseudodrusen (RPD) instance segmentation.",
    "description": "This network detects and segments Reticular Pseudodrusen (RPD) instances in Optical Coherence Tomography (OCT) B-scans which can be presented in a vol or dicom format.",
    "authors": "Yelena Bagdasarova, Scott Song",
    "copyright": "Copyright (c) 2022, uw-biomedical-ml",
    "network_data_format": {
        "inputs": {
            "image": {
                "type": "image",
                "format": "magnitude",
                "modality": "OCT",
                "num_channels": 1,
                "spatial_shape": [
                    496,
                    1024
                ],
                "dtype": "int16",
                "value_range": [
                    0,
                    256
                ],
                "is_patch_data": false,
                "channel_def": {
                    "0": "image"
                }
            }
        },
        "preprocessed_data_sources": {
            "vol_file": {
                "type": "image",
                "format": "magnitude",
                "modality": "OCT",
                "num_channels": 1,
                "spatial_shape": [
                    496,
                    1024,
                    "D"
                ],
                "dtype": "int16",
                "value_range": [
                    0,
                    256
                ],
                "description": "The pixel array of each OCT slice is extracted with volreader and the png files saved to <extracted_dir>/<some>/<file>/<name>/<some_file_name>_oct_<DDD>.png on disk, where <DDD> is the slice number and a nested hierarchy of folders is created using the underscores in the original filename. "
            },
            "dicom_series": {
                "type": "image",
                "format": "magnitude",
                "modality": "OCT",
                "SOP class UID": "1.2.840.10008.5.1.4.1.1.77.1.5.4",
                "num_channels": 1,
                "spatial_shape": [
                    496,
                    1024,
                    "D"
                ],
                "dtype": "int16",
                "value_range": [
                    0,
                    256
                ],
                "description": "The pixel array of each OCT slice is extracted with pydicom and the png files saved to <extracted_dir>/<SOPInstanceUID>/<SOPInstanceUID>_oct_<DDD>.png on disk, where <DDD> is the slice number. "
            }
        },
        "outputs": {
            "pred": {
                "dtype": "dictionary",
                "type": "dictionary",
                "format": "COCO",
                "modality": "n/a",
                "value_range": [
                    0,
                    1
                ],
                "num_channels": 1,
                "spatial_shape": [
                    496,
                    1024
                ],
                "channel_def": {
                    "0": "RPD"
                },
                "description": "This output is a JSON file in COCO Instance Segmentation format, containing bounding boxes, segmentation masks, and output probabilities for detected instances."
            }
        },
        "post_processed_outputs": {
            "binary segmentation": {
                "type": "image",
                "format": "TIFF",
                "modality": "OCT",
                "num_channels": 3,
                "spatial_shape": [
                    496,
                    1024
                ],
                "description": "This output is a multi-page TIFF file. Each page of the TIFF image corresponds to a binary segmentation mask for a single OCT slice from the input volume. The segmentation masks are stacked in the same order as the original OCT slices."
            },
            "binary segmentation overlay": {
                "type": "image",
                "format": "TIFF",
                "modality": "OCT",
                "num_channels": 3,
                "spatial_shape": [
                    496,
                    1024
                ],
                "description": "This output is a multi-page TIFF file. Each page of the TIFF image corresponds to a single OCT slice from the input volume overlayed with the detected binary segmentation mask."
            },
            "instance segmentation overlay": {
                "type": "image",
                "format": "TIFF",
                "modality": "OCT",
                "num_channels": 3,
                "spatial_shape": [
                    496,
                    1024
                ],
                "description": "This output is a multi-page TIFF file. Each page of the TIFF image corresponds to a single OCT slice from the input volume overlayed with the detected binary segmentation mask."
            }
        }
    }
}