dataset_config.py

"""
This module contains information about datasets used in the project.

The datasets are defined as dataset dataclasses.
Each dataset dataclass contains information about the dataset, such as the dataset name, the phases and annotations provided in the source dataset.
It also contains information about how to translate source labels and masks into UMIE format.
"""
from dataclasses import dataclass, field

from config import labels, masks


@dataclass
class MaskColor:
    """A class for storing information about how to recolor a given mask from the source color to the target color."""

    source_color: int
    target_color: int


@dataclass
class DatasetArgs:
    """This class represents required configuration information about each dataset."""

    dataset_uid: str  # unique identifier for the dataset in UMIE
    dataset_name: str  # name of the dataset in UMIE
    phases: dict[
        str, str
    ]  # phase_id is used for encoding the phase in umie img name, phase_name is used for naming the folder
    labels: dict[str, list[dict[str, float]]] = field(
        default_factory=dict
    )  # we allow many-to-many label translations, some labels have multiple RadLex codes
    masks: dict[str, MaskColor] = field(
        default_factory=dict
    )  # mask name and MaskColor object specifying how to recolor the mask


kits23 = DatasetArgs(
    dataset_uid="00",
    dataset_name="kits23",
    phases={"0": "CT"},  # arterial or nephrogenic CT phase (no way to distinguish them in the dataset easily)
    labels={
        "normal": [{labels.NormalityDecriptor.radlex_name: 1}],
        "angiomyolipoma": [{labels.Angiomyolipoma.radlex_name: 1}, {labels.Neoplasm.radlex_name: 1}],
        "chromophobe_rcc": [
            {labels.ChromophobeAdenocarcinoma.radlex_name: 1},
            {labels.RenalAdenocarcinoma.radlex_name: 1},
            {labels.Adenocarcinoma.radlex_name: 1},
            {labels.Neoplasm.radlex_name: 1},
        ],
        "clear_cell_papillary": [
            {labels.ClearCellAdenocarcinoma.radlex_name: 1},
            {labels.PapillaryRenalAdenocarcinoma.radlex_name: 1},
            {labels.RenalAdenocarcinoma.radlex_name: 1},
            {labels.Adenocarcinoma.radlex_name: 1},
            {labels.Neoplasm.radlex_name: 1},
        ],
        "clear_cell_rcc": [
            {labels.ClearCellAdenocarcinoma.radlex_name: 1},
            {labels.RenalAdenocarcinoma.radlex_name: 1},
            {labels.Adenocarcinoma.radlex_name: 1},
            {labels.Neoplasm.radlex_name: 1},
        ],
        "mest": [{labels.Neoplasm.radlex_name: 1}],
        "multilocular_cystic_rcc": [
            {labels.MultilocularCysticRenalTumor.radlex_name: 1},
            {labels.RenalAdenocarcinoma.radlex_name: 1},
            {labels.Adenocarcinoma.radlex_name: 1},
            {labels.Neoplasm.radlex_name: 1},
        ],
        "oncocytoma": [{labels.Oncocytoma.radlex_name: 1}, {labels.Neoplasm.radlex_name: 1}],
        "papillary_rcc": [
            {labels.PapillaryRenalAdenocarcinoma.radlex_name: 1},
            {labels.RenalAdenocarcinoma.radlex_name: 1},
            {labels.Adenocarcinoma.radlex_name: 1},
            {labels.Neoplasm.radlex_name: 1},
        ],
        "rcc_unclassified": [
            {labels.RenalAdenocarcinoma.radlex_name: 1},
            {labels.Adenocarcinoma.radlex_name: 1},
            {labels.Neoplasm.radlex_name: 1},
        ],
        "spindle_cell_neoplasm": [{labels.Neoplasm.radlex_name: 1}],
        "urothelial": [{labels.Neoplasm.radlex_name: 1}],
        "wilms": [{labels.WilmsTumor.radlex_name: 1}, {labels.Neoplasm.radlex_name: 1}],
    },
    masks={
        masks.Kidney.radlex_name: MaskColor(source_color=1, target_color=masks.Kidney.color),
        masks.Neoplasm.radlex_name: MaskColor(source_color=2, target_color=masks.Neoplasm.color),
        masks.RenalCyst.radlex_name: MaskColor(source_color=3, target_color=masks.RenalCyst.color),
    },
)

coronahack = DatasetArgs(
    dataset_uid="01",
    dataset_name="coronahack",
    phases={"0": "Xray"},
    labels={
        "Normal": [{labels.NormalityDecriptor.radlex_name: 1}],
        "PneumoniaBacteria": [{labels.Pneumonia.radlex_name: 1}],
        "PneumoniaVirus": [
            {labels.PneumoniaViral.radlex_name: 1},
            {labels.ViralInfection.radlex_name: 1},
            {labels.Pneumonia.radlex_name: 1},
        ],
    },
)

alzheimers = DatasetArgs(
    dataset_uid="02",
    dataset_name="alzheimers",
    phases={"0": "MRI"},
    labels={
        "MildDemented": [{labels.Dementia.radlex_name: round(2 / 3, 2)}],
        "ModerateDemented": [{labels.Dementia.radlex_name: 1}],
        "NonDemented": [{labels.NormalityDecriptor.radlex_name: 1}],
        "VeryMildDemented": [{labels.Dementia.radlex_name: round(1 / 3, 2)}],
        "mildDem": [{labels.Dementia.radlex_name: round(2 / 3, 2)}],
        "moderateDem": [{labels.Dementia.radlex_name: 1}],
        "nonDem": [{labels.NormalityDecriptor.radlex_name: 1}],
        "verymildDem": [{labels.Dementia.radlex_name: round(1 / 3, 2)}],
    },
)

brain_tumor_classification = DatasetArgs(
    dataset_uid="03",
    dataset_name="brain_tumor_classification",
    phases={
        "0": "T1_weighted_postCM"
    },  # occasionally T2_weighted! (but no way to distinguish them in the dataset easily)
    labels={
        "no_tumor": [{labels.NormalityDecriptor.radlex_name: 1}],
        "glioma_tumor": [{labels.Glioma.radlex_name: 1}],
        "meningioma_tumor": [{labels.Meningioma.radlex_name: 1}],
        "pituitary_tumor": [{labels.Pituitary.radlex_name: 1}],
    },
)

covid19_detection = DatasetArgs(
    dataset_uid="04",
    dataset_name="covid19_detection",
    phases={"0": "Xray"},
    labels={
        "Normal": [{labels.NormalityDecriptor.radlex_name: 1}],
        "BacterialPneumonia": [{labels.Pneumonia.radlex_name: 1}],
        "ViralPneumonia": [
            {labels.PneumoniaViral.radlex_name: 1},
            {labels.Pneumonia.radlex_name: 1},
            {labels.ViralInfection.radlex_name: 1},
        ],
        "COVID-19": [
            {labels.PneumoniaViral.radlex_name: 1},
            {labels.Pneumonia.radlex_name: 1},
            {labels.ViralInfection.radlex_name: 1},
        ],
    },
)

finding_and_measuring_lungs = DatasetArgs(
    dataset_uid="05",
    dataset_name="finding_and_measuring_lungs",
    phases={"0": "CT"},
    masks={masks.Lung.radlex_name: MaskColor(source_color=255, target_color=masks.Lung.color)},
)

brain_with_intracranial_hemorrhage = DatasetArgs(
    dataset_uid="06",
    dataset_name="brain_with_intracranial_hemorrhage",
    phases={"0": "Bone", "1": "Brain"},
    labels={
        "brain_hemorrhage": [{labels.Hemorrhage.radlex_name: 1}],
        "normal": [{labels.NormalityDecriptor.radlex_name: 1}],
    },
    masks={masks.Hemorrhage.radlex_name: MaskColor(source_color=255, target_color=masks.Hemorrhage.color)},
)

lits = DatasetArgs(
    dataset_uid="07",
    dataset_name="lits",
    phases={"0": "CT"},
    labels={
        "Neoplasm": [{labels.Neoplasm.radlex_name: 1}],
        "NormalityDescriptor": [{labels.NormalityDecriptor.radlex_name: 1}],
    },
    masks={
        masks.Liver.radlex_name: MaskColor(source_color=150, target_color=masks.Liver.color),
        masks.Neoplasm.radlex_name: MaskColor(source_color=145, target_color=masks.Neoplasm.color),
    },
)

brain_tumor_detection = DatasetArgs(
    dataset_uid="08",
    dataset_name="brain_tumor_detection",
    phases={"0": "MRI"},
    labels={
        "Y": [{labels.NormalityDecriptor.radlex_name: 1}],
        "N": [{labels.Neoplasm.radlex_name: 1}],
    },
)

knee_osteoarthritis = DatasetArgs(
    dataset_uid="09",
    dataset_name="knee_osteoarthritis",
    phases={"0": "CT"},
    labels={
        "0": [{labels.NormalityDecriptor.radlex_name: 1}],
        "1": [{labels.Osteoarthritis.radlex_name: round(0.25, 2)}],
        "2": [{labels.Osteoarthritis.radlex_name: round(0.5, 2)}],
        "3": [{labels.Osteoarthritis.radlex_name: round(0.75, 2)}],
        "4": [{labels.Osteoarthritis.radlex_name: 1}],
    },
)

brain_tumor_progression = DatasetArgs(
    dataset_uid="10",
    dataset_name="brain_tumor_progression",
    phases={
        "0": "T1post",
        "1": "dT1",
        "2": "T1prereg",
        "3": "FLAIRreg",
        "4": "ADCreg",
        "5": "sRCBVreg",
        "6": "nRCBVreg",
        "7": "nCBFreg",
        "8": "T2reg",
    },  # masks based on T1 weighed
    labels={},
    masks={masks.Neoplasm.radlex_name: MaskColor(source_color=255, target_color=masks.Neoplasm.color)},
)


chest_xray14 = DatasetArgs(
    dataset_uid="11",
    dataset_name="chest_xray14",
    phases={"0": "Xray"},
    labels={
        "No Finding": [{labels.NormalityDecriptor.radlex_name: 1}],
        "Atelectasis": [{labels.Atelectasis.radlex_name: 1}],
        "Cardiomegaly": [{labels.BoxlikeHeart.radlex_name: 1}],
        "Effusion": [{labels.PleuralEffusion.radlex_name: 1}],
        "Infiltration": [{labels.Consolidation.radlex_name: 1}],
        "Mass": [{labels.Lesion.radlex_name: 1}, {labels.Mass.radlex_name: 1}],
        "Nodule": [{labels.Lesion.radlex_name: 1}],
        "Pneumonia": [{labels.Pneumonia.radlex_name: 1}],
        "Pneumothorax": [{labels.Pneumothorax.radlex_name: 1}],
        "Consolidation": [{labels.Consolidation.radlex_name: 1}],
        "Edema": [{labels.PulmonaryEdema.radlex_name: 1}],
        "Emphysema": [{labels.Emphysema.radlex_name: 1}],
        "Fibrosis": [{labels.Fibrosis.radlex_name: 1}],
        "Pleural_Thickening": [{labels.Thickening.radlex_name: 1}],
        "Hernia": [{labels.Hernia.radlex_name: 1}],
    },
)

coca = DatasetArgs(
    dataset_uid="12",
    dataset_name="coca",
    phases={"0": "CT"},
    masks={masks.CalciumScore.radlex_name: MaskColor(source_color=20, target_color=masks.CalciumScore.color)},
)

brain_met_share = DatasetArgs(
    dataset_uid="13",
    dataset_name="brain_met_share",
    phases={
        "0": "T1_weighted_preCM_spin-echo_pre-contrast",
        "1": "T1_weighted_postCM",  # This one was used to generate the masks
        "2": "T1_gradient_echo_postCM",  # using an IR-prepped FSPGR sequence
        "3": "T2_FLAIR_postCM",
    },
    masks={masks.Metastasis.radlex_name: MaskColor(source_color=255, target_color=masks.Metastasis.color)},
)

cmmd = DatasetArgs(
    dataset_uid="18",
    dataset_name="cmmd",
    phases={"0": "MG"},
    labels={
        "calcification": [{labels.Calcification.radlex_name: 1}],
        "mass": [{labels.Mass.radlex_name: 1}, {labels.Lesion.radlex_name: 1}],
        "both": [{labels.Calcification.radlex_name: 1}, {labels.Mass.radlex_name: 1}, {labels.Lesion.radlex_name: 1}],
        "Benign": [{labels.Benign.radlex_name: 1}],
        "Malignant": [{labels.Malignant.radlex_name: 1}],
    },
)
all_datasets = [obj for name, obj in globals().items() if isinstance(obj, DatasetArgs)]