Mimic4Dataset.py

import csv
import json
import os
import pandas as pd
import datasets

_DESCRIPTION = """\
Dataset for mimic4 data, by default for the Mortality task.
Available tasks are: Mortality, Length of Stay, Readmission, Phenotype.
The data is extracted from the mimic4 database using this pipeline: 'https://github.com/healthylaife/MIMIC-IV-Data-Pipeline/tree/main'
#TODO ADD DESCRIPTION COHORTS
"""

_HOMEPAGE = "https://huggingface.co/datasets/thbndi/Mimic4Dataset"
_CITATION = "https://proceedings.mlr.press/v193/gupta22a.html"

class Mimic4Dataset(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="Phenotype",
            version=VERSION,
            data_dir=os.path.abspath("./data/csv/Phenotype"),
            description="Dataset for mimic4 Phenotype task",
        ),
        datasets.BuilderConfig(
            name="Readmission",
            version=VERSION,
            data_dir=os.path.abspath("./data/csv/Readmission"),
            description="Dataset for mimic4 Readmission task",
        ),
        datasets.BuilderConfig(
            name="Length of Stay",
            version=VERSION,
            data_dir=os.path.abspath("./data/csv/Lenght_of_Stay"),
            description="Dataset for mimic4 Length of Stay task",
        ),
        datasets.BuilderConfig(
            name="Mortality",
            version=VERSION,
            data_dir=os.path.abspath("./data/csv/Mortality"),
            description="Dataset for mimic4 Mortality task",
        ),
    ]

    DEFAULT_CONFIG_NAME = "Mortality"

    def _info(self):
        

        features = datasets.Features(
            {
                "gender": datasets.Value("string"),
                "ethnicity": datasets.Value("string"),
                "insurance": datasets.Value("string"),
                "age": datasets.Value("int32"),
                "COND": datasets.Sequence(datasets.Value("int32"), length=None),
                "MEDS": datasets.Sequence(datasets.Value("int32"), length=None),
                "PROC": datasets.Sequence(datasets.Value("int32"), length=None),
                "CHART": datasets.Sequence(datasets.Value("int32"), length=None),
                "OUT": datasets.Sequence(datasets.Value("int32"), length=None),
                "label": datasets.ClassLabel(names=["0", "1"]),
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        data_dir = self.config.data_dir

        # Collect file paths for all CSV files in the subfolders
        train_files = []
        for split_name in os.listdir(data_dir):
            split_dir = os.path.join(data_dir, split_name)
            if os.path.isdir(split_dir):
                for file_name in os.listdir(split_dir):
                    if file_name.endswith(".csv"):
                        file_path = os.path.join(split_dir, file_name)
                        train_files.append(file_path)

        # Return a single SplitGenerator for the train split
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "filepaths": train_files,
                    "split": datasets.Split.TRAIN,
                },
            )
        ]


    def _generate_examples(self, filepaths, split):
        #each 3 successive files are the same admission (demographic, static, dynamic)
        labels = pd.read_csv("./data/csv/"+self.config.name +"labels.csv")
        for i in range(0, len(filepaths), 3):
            file1, file2, file3 = filepaths[i:i+3]
            static_file = file1 if "static.csv" in file1 else file2 if "static.csv" in file2 else file3 
            demographic_file = file1 if "demo.csv" in file1 else file2 if "demo.csv" in file2 else file3
            dynamic_file = file1 if "dynamic.csv" in file1 else file2 if "dynamic.csv" in file2 else file3 
            
            #dynamic
            dyn = pd.read_csv(dynamic_file, header=[0, 1])
            meds = dyn['MEDS']
            proc = dyn['PROC']
            chart = dyn['CHART']
            out = dyn['OUT']
            
            #static
            stat = pd.read_csv(static_file, header=[0, 1])
            stat = stat['COND']
            
            #demo
            demo = pd.read_csv(demographic_file, header=0)

            #dict
            stat_dict = stat.iloc[0].to_dict()
            demo_dict = demo.iloc[0].to_dict()
            meds_dict = meds.iloc[0].to_dict()
            proc_dict = proc.iloc[0].to_dict()
            chart_dict = chart.iloc[0].to_dict()
            out_dict = out.iloc[0].to_dict()

            #get stay_id which is the name of the folder containing the files
            stay_id = demographic_file.split("/")[-2]
            #get the label
            label = int(labels.loc[labels['stay_id'] == stay_id]['label'])

            yield stay_id, {
                "gender" : demo_dict['gender'],
                "ethnicity" : demo_dict['ethnicity'],
                "insurance" : demo_dict['insurance'],
                "age" : demo_dict['age'],
                "MEDS" : meds_dict,
                "PROC" : proc_dict,
                "CHART" : chart_dict,
                "OUT" : out_dict,
                "COND" : stat_dict,
                "label" : label
            }